NailScience — Brand Database (BRAND.md)
The single source of truth for every durable fact about this brand. Any note David sends
(inventory, suppliers, accounts, contacts, decisions) is filed HERE the same turn it arrives —
before any context compaction. Memory files only POINT here. Never store secrets (tokens, keys,
passwords): reference the env file that holds them. Keep entries dated. Hub renders this file.
Status 2026-10-13: RELAUNCH — kickoff on the new playbook. Retired 2026-06-15; David 2026-10-13: "We will be relaunching trynailscience.com. We will be adopting the new playbook." The existing folder and site are the base — CONFIRMED (David 2026-10-14: "yes"). The index and the plan: briefs/nailscience/RELAUNCH.md. Console lane: nailscience — 7 of 8 kickoff asks answered 2026-10-14 (§6); ns-webhook-crossfire still open.
0 · Identity & entities
| Field | Value |
| Legal entity / owner | NailScience Co. (the name on the logo and the adscalerads client "NailScience Co.", id 1) — legal entity TO CONFIRM |
| Key people & roles | David Sanghera — owner / operator. "Dr. Franks" — the house doctor persona (podiatrist voice of the advertorial and the emails); a composite, never a real person (CL-14), never reports lab results |
| Support email / phone | support@trynailscience.com (the Klaviyo sender) · phone TO FILL |
| Brand relationship | own brand |
| Teams | Netlify team EcomScalers · Meta BM "DeltaSift" (Liaison – A) in May 2026 |
1 · Products & SKUs
| Code | Product | Size / format | UPC | Formula / spec source | Supplier |
| NS-01 | 2-Step Nail Restoration Kit — 3 pens = one month. Step 1 Nail Prep pen (urea + lactic acid + propylene glycol, the softener). Step 2 Maximum-Strength Antifungal pen (25% undecylenic acid + tea tree, oregano, manuka, clove oils) | 3 ml PP pen 1.1 × 12.5 cm · PE bag 8 × 18 cm · carton 56 × 36 × 21 cm | TO FILL | copy brief tools/deep-researcher/runs/nailscience/at-ns/07_copy-brief/response.md · the live PDP | Lansil Global (Shenzhen Lansil Trading Ltd, support@lansilglobal.com) |
✅ FORMULA — CLOSED (David 2026-10-14: "What's on the PDP is what's in the pens"). The pens carry the PDP formula above; the salicylic-acid line on the Lansil quote was the factory's reference formula, not our fill. The Truth Gate's first claim is settled. Undecylenic acid is an FDA OTC antifungal monograph active (21 CFR 333): establishment registration, NDC and a Drug Facts panel are the officer's questions, not a constraint on the first draft (CL-00).
29next store items (acm2): 3 "NailScience™ 2-Step Nail Restoration Kit" · 4 "Fungus Destroyer Pack" · 37 Shipping Protection · 70 Extra Strength. Packages: see §3.
2 · Inventory & supply chain
| SKU | Units on hand | Location / 3PL | Lot / mfg date | Shelf life | COGS (landed) | MOQ / reorder lead time | Updated |
| NS-01 | Stock on hand in Shenzhen — enough to relaunch (David 2026-10-14: "Ignore, we have stock in Shenzhen"); units not counted | 4PX drop-ship from Lansil's Shenzhen warehouse (no US stock) — the May model stands for the relaunch. Lansil Las Vegas 3PL ($1.00 pick/pack) was modelled, never confirmed | — | — | 3-pen kit $1.97 product + $0.30 pick/pack + $5.44 4PX postage = $7.71 to the customer · 6-pen $9.88 · 9-pen $17.59 (ships as two parcels) | 3,000 MOQ · quote valid to 2026-05-16 (expired — re-quote) | 2026-10-13 |
Notes: quote PDF + screenshot in logistics/nailscience/quotes/; cost models (drop-ship / sea→Vegas / air→Vegas), scaling tiers and the supplier question list in logistics/nailscience/scenarios.json. Lansil Global is also a consulting client (clients/lansilglobal/).
3 · Offer & pricing
The relaunch offer — LOCKED 2026-10-14 (David: "Ladder A is good")
| Row (data-qty) | Customer name | Kits · pens | Price | Compare-at | Per kit | Lane |
| 1 | One Month Supply — 1 Kit (3 pens) | 1 · 3 | $29.99 | $49.99 | $29.99 | Shopify ladder |
| 3 | Buy 2 + Get 1 FREE — 3 Kits (9 pens) · MOST POPULAR · default | 3 · 9 | $59.98 | $149.97 | $19.99 | Shopify ladder |
| 5 | Buy 3 + Get 2 FREE — 5 Kits (15 pens) · BEST DEAL | 5 · 15 | $89.97 | $249.95 | $17.99 | Shopify ladder |
| — | Monthly (subscription) | 1 · 3 | $24.99 / 30 d | — | $24.99 | 29next |
| — | VIP (OTO 1, text-only) · Extra Strength (OTO 2) · Shipping Protection (bump) | $14.99 → $9.99/mo · $11.99 · $9.99 | 29next, carried over from May pending the OTO review |
The BOGO formula (3 kits = 2× the single, 5 kits = 3× the single); compare-at = $49.99 × kits (the May strike per kit). The default row is 3 kits — the three months one nail cycle needs. Guarantee 30-Day Money-Back. Free US shipping.
Economics gate on Ladder A (2026-10-14, Shenzhen drop-ship costs, even tier mix): FIX at a $60 CPA — the average first order $59.98 leaves $40.80, 32% of buyers would have to reorder; the monthly lane pays back only on optimistic retention; PASSES at a CPA under $43.16. David built it anyway ("Ladder A is good") — $43 is the CPA to beat in Phase 7; the gate re-runs on live data in Phase 8.
#### Where Ladder A and the trial live on 29next — campaign 1494 "NailScience Co." on acm2 (built 2026-10-14 late; products/nailscience/build_29next.py owns every id, map = creative/nailscience/29next-map.json)
| Package | Rides | Price | Role | ||
2 NailScience — Mild Fungus (1 kit) (was Mild Issue (1 kit) 2026-10-17 → 10-25, 1 Kit (3 pens) before) | product 3 / variant 2 (sku nailscience) — the May pkg 2 renamed in place | $24.99 (since 2026-10-20; $29.99 from the relaunch to then) | Ladder A one-kit row (a stray price_recurring 9.99 sits on it, inert — is_recurring false) | ||
10 NailScience — Moderate Fungus (3 kits) (was Moderate Issue 10-17 → 10-25) (was … Buy 2 + Get 1 FREE (3 kits)) | product 202 / variant 203 (sku nailscience-3kits) | $59.98 | Ladder A default row | ||
11 NailScience — Severe Fungus (5 kits) (was Full Treatment 10-17 → 10-25) (was … Buy 3 + Get 2 FREE (5 kits)) | product 204 / variant 205 (sku nailscience-5kits) | $89.97 | Ladder A best value | ||
12 Try Before You Buy — Monthly | product 206 / variant 207 (sku nailscience-trial-monthly) | $0.00 today · $24.99 on day 14 · then $24.99 every 30 days (the order webhook flips the cadence) | the trial row / OB-C subscribe lane (plan=tbyb-sub) | ||
13 Try Before You Buy — One Kit | product 208 / variant 209 (sku nailscience-trial-one) | $0.00 today · $29.99 on day 14 · nothing renews (the webhook cancels it) | the one-time lane (plan=tbyb-one, a checkout deep link) | ||
| 6 Shipping Protection $9.99 (bump) · 7 VIP $14.99 → $9.99/mo · 8 Fungus Destroyer Pack $29.99 (OTO) · 9 Extra Strength Upgrade $11.99 (OTO) | kept as the May build had them — the bump CONTAINER ships hidden since 2026-05-12 (8ed06a76c, "hide + uncheck Shipping Protection bump for higher conversion rate"; the May call stands, the trial's deselect is the guard behind it); the VIP books as a month/1 $9.99 subscription; the VIP PAGE rebuilt 2026-10-15 (§6) | ||||
| 3 / 4 / 5 — the May 2/3/4-kit packages (qty 2/3/4 on variant 2) | legacy, unlinked; a stale `?pkg=3 | 4 | 5` is rewritten to 10/10/11 by the checkout's head shim |
Rules the store taught on 2026-10-14 (why the shape above): one package per product variant per campaign ("Product variant [2] already exists in another package"), product 3 holds exactly one child (no variant attributes — POST /products/3/variants/ UPSERTS that child: it rewrote variant 2's sku/price for a minute before the restore; kit_variant() now self-heals), a package create ignores prices (it comes back at the variant's retail — the builder PATCHes right after), no settable quantity on the admin API, a second attach of the same shipping code is refused and the admin key may not create store-level shipping methods (403) → the store-level method was created in the 29next admin UI on 2026-10-15 22:52 PT (in David's Chrome, on his word: Settings → Fulfillment → Add Shipping Method — "Try Before You Buy Shipping", code tbyb, USD $6.99 per order, everywhere, public; the store's only other method is default) and attached to campaign 1494 by build_29next.py --create as shipping ref 2 @ $6.99 (ref 1 default @ $0.00 stays the ladder's) — the map carries tbyb_shipping_ref: 2, the checkout's trial card data-next-shipping-id="2". Webhook env on the production site (set 2026-10-14 late, set_prod_env.py --apply): NEXT_ADMIN_BASE · NEXT_ADMIN_API_KEY · ORDER_SKU_PREFIX=nailscience (ends the Charm/CellScience crossfire the moment the relaunch build deploys to production) · TBYB_LANES keyed by the two trial package NAMES · TBYB_FULFILL_AS/TAG.
#### The Shopify ladder — shop.trynailscience.com (LIVE 2026-10-15 → DORMANT since 2026-10-21: David, "lets make the checkout like it was on 29next, not shopify" — the store, product, channel and tokens stay; handoff.on: false in the spec; ?co=shopify previews)
David delivered our own app's credentials on 2026-10-15 ("NailScience Checkout" → ~/.shopify/nailscience.env, mode 600, never in git). The house kit ran end to end: tools/shopify_store.py nailscience (read → --create → --shipping → --storefront-token) → tools/shop_handoff.py on the PDP (through build_pdp_extras.py, last before </body>), the home and the collection; spec creative/nailscience/shopify-spec.json, map creative/nailscience/shopify-map.json (no secrets).
| Store | NailScience · d65bq7-ih.myshopify.com · primary domain shop.trynailscience.com (DNS → shops.myshopify.com, done) · USD · plan Shopify · Shopify Payments ON (US) · Admin API 2026-07, every scope the kit needs |
| Open on David's side | time zone Asia/Manila → America/Los_Angeles (Settings → General) · the Facebook & Instagram and Google & YouTube channels are not on the store (channels today: Online Store, POS, Shop) — ask ns-shopify-finish |
| Product | nailscience-2-step-nail-restoration-kit (10590520705313) "NailScience 2-Step Nail Restoration Kit", option Package — 1 Kit $29.99 / $49.99 · variant 53798793380129 · NS-01_1 (← pkg 2) · 3 Kits $59.98 / $149.97 · 53798793412897 · NS-01_3 (← pkg 10) · 5 Kits $89.97 / $249.95 · 53798793445665 · NS-01_5 (← pkg 11); Online Store only; description = David's subhead + "Three pens per kit · one kit a month · 30-Day Money-Back Guarantee"; image /assets/bundle-1.jpg |
| Shipping | one Worldwide zone (237 countries): Free Shipping on any order ≥ $1 · Standard $6.99 only on a $0 order (the house guard; no gift at launch) |
| Handoff | LIVE on the PDP, home and collection (tracking/shop-handoff.js): a ladder row's checkout link → the Shopify cart permalink carrying UTMs / click ids / _fbp / _fbc as attributes; Buy with Shop Pay under #atcBtn follows the row; ?pkg=2/10/11 → tier; trial lanes (plan=tbyb-*, pkgs 12/13) and VIP stay on 29next; ?co=29next keeps a visit on the 29next checkout; storefront token nailscience-pdp |
| Proof | shop_handoff_check.py 24/24 offline · shopify_matrix.py --quick on staging 16/16 (iPhone WebKit + desktop Chromium: T1–T3 every tier → Shopify with Shop Pay in the express row · P1 Buy with Shop Pay · P2 hidden on the trial row · G1 trial → 29next · H1 ?co=29next · Z zero page errors; tools/qa/results/nailscience-shopify-20261015-0934) |
The May 2026 offer (history — what the relaunch replaced)
| Item | 29next pkg | Price | Strike | Per month |
| 1 Month Supply / 3 pens ("Mild Issue") | 25289 | $21.99 | $49.99 | $21.99 |
| 2 Month Supply / 6 pens ("Moderate Issue", default) | 25290 | $39.98 | $99.98 | $19.99 |
| 3 Month Supply / 9 pens ("Full Treatment") | 25291 | $53.97 | $129.99 | $17.99 |
| Shipping Protection (order bump) | 25522 | $9.99 | 3 of 26 May orders | |
| VIP (OTO 1, text-only rule applies) | 25765 | $14.99 then $9.99/mo | 7 of 26 May orders | |
| Extra Strength upgrade (OTO 2) | 26961 | $11.99 | 7 of 26 May orders | |
Bundle OTO (/special-offer/bundle/) | TO FILL |
- Guarantee — LOCKED (David 2026-10-14: "yes"): 30-Day Money-Back Guarantee on every surface. The live PDP and checkout still say 90-Day and the Klaviyo drafts 60-day — every page, badge, legal clause and email moves to 30 in the rebuild. Free US shipping on the PDP.
- Checkout — LOCKED (David 2026-10-14): the Shopify ladder (shop.trynailscience.com, our own app) with 29next only for the trial lane, VIP and the upsell chain → REVERSED 2026-10-21 (David: "lets make the checkout like it was on 29next, not shopify"): every lane checks out on our 29next checkout; the Shopify handoff is dormant (§6) — the house default. Today's checkout is 29next on the shared store; no subscription or trial lane existed.
- Economics gate 2026-10-13 on the May offer (May costs, estimates): KILL at a $60 CPA and at the May $62.98. One-time: the average order $38.65 leaves $25.22; 58% of buyers would have to reorder. Passes at a CPA under $35.30. David 2026-10-14: "Rebuild the offer first, then re-run the gate" → three ladders on the house BOGO formula (3 kits = 2× single, 5 kits = 3× single; monthly lane ≈ single − 15%) ran through the gate 2026-10-14 with the same Shenzhen costs and an even tier mix (no orders yet) — ask
ns-offer-pick:
| Candidate | 1 kit | 3 kits | 5 kits | Monthly | Gate @ $60 | Passes under | Avg first order leaves |
| A | $29.99 | $59.98 | $89.97 | $24.99 | FIX (32% must reorder) | $43.16 | $40.80 |
| A, since 2026-10-20 | $24.99 | $59.98 | $89.97 | $24.99 | FIX (35% must reorder) | $43.16 | $39.27 |
| B | $34.99 | $69.98 | $104.97 | $29.99 | FIX (17% must reorder) | $56.26 | $50.01 |
| C | $39.99 | $79.98 | $119.97 | $34.99 | PASS (monthly pays back by month 6; first order ≈ break-even) | $69.36 | $59.22 |
Reference: Crystal Flush / Lunavia / Kerassentials retail $69–$140 per month (copy brief Q17). David picked A (2026-10-14, in chat: "Ladder A is good"). tools/proforma/assumptions/nailscience.json now carries Ladder A (the scratch a/b/c files were deleted); page /nailscience/economics/.
4 · Accounts & IDs (no secrets)
| System | ID / handle | Notes | ||||||||
| Domain(s) / DNS | trynailscience.com (apex; www 301 → apex) · quiz.trynailscience.com | thehealthreportdaily.com hosted the Dr. Franks advertorial + NS-P2 until 2026-07, then went to the bronze prelanders (commit 84aea4c93); both NS pages are dark → since 2026-10-16 both are served there again at their April URLs (/ = NS-P3, /blog/ranking/ = NS-P2) by tools/deploy-health-report.sh (the SkinScience hub site that owns the domain); their server events go to trynailscience.com/api/capi | ||||||||
| Netlify sites (EcomScalers) | trynailscience.com = vocal-sunburst-5f0620 · id dd6c30db-30da-4e7f-827d-ff0a12426637 — published deploy 2026-05-20, 2 functions · quiz = wonderful-meerkat-49cae2 · id b7ddc126-97b7-44a0-a457-85a772de5982 (2026-05-17) · Dr. Franks / NS-P2 = dreamy-licorice-eedba8 · id a2bbc9e1-cf74-4df9-937d-2336b8b140fa (no custom domain) | Netlify env on the main site: FB_PIXEL_ID · FB_CAPI_ACCESS_TOKEN · GA4_MEASUREMENT_ID · GA4_API_SECRET · NEXT_WEBHOOK_SECRET. No ORDER_SKU_PREFIX. No deploy script yet — the site was last deployed bare | ||||||||
| 29next store / campaign | acm2.29next.store — NailScience's store, SHARED with Charm (campaign 1877, since 2026-09-04) and CellScience (2026-09). Admin key ~/.29next/nailscience.env; the NS campaign key sits in products/nailscience/checkout/nailscience/config.js | Webhook 67 "Order Created" → trynailscience.com/api/order-webhook is ACTIVE (checked 2026-10-13) and the live May-20 function has no brand gate → every Charm and CellScience order fires a NailScience Purchase (filed on Charm as charm-ns-webhook-crossfire). Ask ns-webhook-crossfire | ||||||||
| Meta BM · ad account · pixel | pixel 1341229411229569 · ad account act_832705320684189 (Liaison – A, DeltaSift BM) — CONFIRMED the same for the relaunch (David 2026-10-14: "Yes same.") | CAPI token → Netlify env only. Pattern I ported 2026-05-16; Pattern K (wait for _fbp/_fbc) rolled to live brands 2026-07 — NS was dead, verify · the adscaler system user asu (968334878869352) holds DRAFT / ANALYZE / ADVERTISE / MANAGE here since 2026-10-16 (David assigned it; it had none, so every switch was refused) | ||||||||
| Facebook Page (the ads) · April campaigns | Page Carol Anne Foster 927552797118674 — every April ad posts from it (the ads' tracking specs) · J&T – Foot Doctor Reveals 120243267906260510 · J&T – Top 5 Rankings 120243267753990510 · J&T – 5 Reasons 120243267606940510 (CBO $50 each, PAUSED since May; the ad sets inside are still ACTIVE) | Media plan 2026-10: the four switch-ons (W1 W2 W3 + W4/AD025) go back ON in those campaigns; briefs/nailscience/ad-codes.json backfilled 2026-10-14 (35 codes, nothing renamed on Meta) | ||||||||
| People on the ad account | David Sanghera · asu (adscaler's system user, every launch and switch of ours) · Fotini Panagiotaropoulou (Power Editor) | Read from the activity log 2026-10-30: since 2026-10-19 Fotini paused round one's two campaigns (10-19 07:44 PT), renamed J&T – Top 5 Rankings → 'Prospecting \ | US \ | ALL 7DC WINNERS \ | Ranking page' and ran it 10-22 → 10-26 (W3 / W4 edited and re-reviewed 10-20 / 10-22), and on 10-29 13:02 PT created 'Prospecting \ | US \ | Testing \ | CBO' ($150, 8 image ads) + 'Prospecting \ | US \ | BFCM' ($100, 3) on the Top 5 page — outside the media plan. 2026-11-01 01:18 PT: she cut Testing CBO $150 → $30/day and switched BFCM off. Read the activity log before quoting the account's daily total. |
| GA4 property · measurement id · GTM | GA4 G-HND2VE2X2S (API secret in Netlify env) · GTM GTM-P9MSJBP7 on PDP, checkout and both prelander sites — kept as-is (David 2026-10-14: "Yes same.") | Charm runs without GTM; NailScience keeps its container | ||||||||
| Klaviyo org | Nail Science · public RA2aJd · key ~/.klaviyo/nailscience.env · lists Y4EDQu (Email, 54) / SAKJXD / V6PVs9 · flows DRAFT: Welcome Yj86A3 · Cart U2C2B4 · Leads WTPuDZ · sender Dr. Franks <support@trynailscience.com> | Bodies live only in Klaviyo (the 9 HTML bodies were never committed); activation gates in the imported record below; no ongoing 29next→Klaviyo sync | ||||||||
| adscalerads client · Airtable base | client id 1 "NailScience Co." · Airtable base TO FILL | |||||||||
| Cookie / event prefix | _ns_ (_ns_eid, _ns_utms, _ns_landing, _ns_referrer) | also used by HumanFrame (legacy fork of this template) | ||||||||
| Profit / MER · Stripe | profit/nailscience/config.json (Stripe default 2.9% + $0.30) | May 7–31 Meta + orders pulled |
5 · Compliance ceiling (the officer's inputs — never a constraint on the first draft, CL-00)
- Category: OTC topical antifungal active (undecylenic acid, FDA OTC monograph 21 CFR 333) sold as cosmetic nail restoration ("not a cure" positioning in the copy brief).
- Officer files
briefs/nailscience/compliance.md+forbidden-terms.txt— NOT WRITTEN (Phase 0 output). - Meta health & wellness category restrictions apply per domain; the May campaigns optimised for Purchase.
- Proof lines in the old copy to verify at the Truth Gate: "40,000+ customers", "3,214 Trustpilot reviews", "doctor-trusted".
6 · Decisions & taste log (dated, David's verdicts = ground truth)
- 2026-11-03 — The media plan lives on Media Plan HQ, as the Funnel map (David: "a simple spreadsheet doesn't work" → four template lanes → "lane D." → "lets make a mediaplan-hq.netlify where we house these … update all the reference links and remove all the old boards"; the schedule board: "Move it into Media Plan HQ"): https://mediaplan-hq.netlify.app/nailscience/ · its schedule board https://mediaplan-hq.netlify.app/nailscience/schedule/. Each round is the path a buyer takes (targeting → budget → ads → landing page → goal) with the counts on every step and the 30% page bar on the drop; the plan page names no other brand. Retired: the 2026-11-02 sheet (Copy Sheet HQ's Media plan tab), the Launch Center plan page and board (their addresses forward to the HQ), the design-lane boards.
- 2026-10-18 — The natives pass boarded (r1) — the Native Ads Research SOP (Anirudh's doc, David: "we need to somehow integrate this into our angle studio and research") now runs inside the Angle Studio (
skills/angle-studio/SKILL.md §3b,tools/angle-studio/natives.py). nail-fungus: 84 unique copies → 81 long-copy image ads pass the stack (≥1,500 chars, ≥14 days, active) · 6 in both sorts (longest-running ∩ most-reused) · 38 hit the health keyword set; from a fresh Apify sweep of 'toenail fungus' with the SOP's filters in the Ad Library URL (230 ads, ≈ $2.40). Board: https://launch-center-hq.netlify.app/nailscience/boards/nailscience--natives-r1/ (Angle Studio station → Natives r1): David reads the first five lines of each out loud, KEEP if it pulls before the product is clear, SKIP if flat. Both-lists leaders: Fungus Free Reset 1679682190824033 (38d, 22,388 chars); Fungus Free Reset 2599841497160325 (38d, 25,408 chars); Fungus Free Reset 1883390633067106 (38d, 11,295 chars). A kept native is a structure reference only (CL-13);natives.py pullthendistillranks the keeps first in every angle's evidence and the packets carry their leads. OPEN: impression rank is not in the Ad Library (a spy tool would add it — his call); the 2–3×/week cadence is ~$2 a sweep on Apify (his call). - 2026-10-18 (8:40 PM PT) — "set each nailscience campaign to $50 budget so it only sepnds $100 per day total" (David). Both campaigns were already at $50 a day (CBO, unchanged since the switch-on); nothing else is on. The overspend is Meta's budget flexibility: up to 75% over on a day ($87.50 a campaign), never more than 7× the budget in a Sunday–Saturday week. Oct 17: $85.21 + $86.04 = $171.25; Oct 18 to 8:37 PM: $155.08. Since the switch-on: $336.35, 2,586-ish impressions, 17 checkouts started, 0 orders (Meta, Shopify and 29next all show none). By ad set: W1 $51.15 (1 checkout) · W2 $112.92 (3) · W3 $17.81 (0) · W4 $154.47 (13, $11.88 each, CPM $126). Day-3 rule (Oct 19): W2 and W4 past $60 with no order → off; W1 and W3 one more day. A hard $100/day needs a budget / spending-limit route in adscaler (it has toggle, launch, sync only); options put to David: daily spending limits topped up at midnight, or ~$28.50 budgets.
- 2026-10-16 (night) — THE FOUR APRIL WINNERS ARE BACK ON, 10:41 PM PT (David: "done", after assigning the adscaler system user asu Manage campaigns on Liaison – A in DeltaSift Business settings; asu now holds DRAFT / ANALYZE / ADVERTISE / MANAGE there).
tools/ad_codes.py on … --yes: 10 of 10 switches 200 — seven sibling ad sets paused, W4's ad set ACTIVE,J&T – Foot Doctor RevealsandJ&T – Top 5 RankingsACTIVE at $50 each. Read back from Meta: only W1 / W2 / W3 / W4's ad sets are on, each with only its winner, all four ads ACTIVE/ACTIVE (no review). $100 a day; day 1 = Oct 17 (LA time), day 2 rules Oct 18, day 3 keep/off Oct 19. The third $50 is its own ask (ns-third-slot: V12 + two more videos, or seven kept statics). - 2026-10-16 (night) — THE GO, IN CHAT; THE SWITCH-ON REFUSED BY META, NOTHING CHANGED (David: "For NailScience, can we turn on the ads that i was talking about, now that the prelanders are restored and the new site is pushed?" then "do the prelanders have the updated pdp url on them? for nailscience"). Preflight PASSED: both April pages' buttons go to
trynailscience.com/products/nail-restoration-kit/(6 on the Foot Doctor page, 5 on the Top 5 page, the sticky bars included;tools/qa/nailscience_prelander_proof.py41/41 live, desktop + iPhone); the four ads are ACTIVE at the ad level with no review feedback and no issues (W1 W2 → thehealthreportdaily.com/, W3 W4 → /blog/ranking, which 301s to /blog/ranking/ keeping the fbclid); the brand site's last deploy (09:53 PT) came before its tracking proofs.tools/ad_codes.py on … --yesstopped at toggle 1 of 10: "(#200/2490591) To create or edit ads for this ad account, contact an admin to get Advertise permission". The adscaler system user asu (968334878869352) holds NO tasks on Liaison – A (on Liaison – B, LL1 and RS – Spare it holds DRAFT / ANALYZE / ADVERTISE / MANAGE). Fix = David in DeltaSift Business settings: assign asu to Liaison – A with Manage campaigns; then the same command, unchanged. The go covers the four April ads (two campaigns, $100 a day); the third $50 (V12 + two more videos, or seven kept statics) is still his call. - 2026-10-16 — Rounds take a new shape: seven statics in one ad set, videos three to an ad set (David: "So lets modify the rules of the rounds. 7 Statics in One Adset. 3 video Ads per Adset. This is how we can test more and get better data. Lets use this moving forward."), and a page has to send more than 30% of its readers on to the product page (David: "in the round system, in the prelanders need a over > 30% click through rate from LP to the PDP. That should be included in this, because we need to switch rounds if the LP has low CTR, then we know its a LP problem"). W1–W4 stay in their own campaigns, switched on untouched; V12 can't run alone, so two more videos join its ad set or the third $50 runs seven of the 54 kept statics (his call with the go).
- 2026-04-03 — NailScience Co. branding + 3-pen gallery; PDP at
/, checkout at/checkout/live 2026-04-11. - 2026-04-25 → 04-28 — J&T creative rounds 1–6: 81 images, 54 KEEP / 27 KILL (
tools/jt/judgement-nailscience-2026-04-28-225837.json). Kill reasons, law for the next round: "toe nail needs to be more gross looking" · "hard to see the nail" · "too playful and fake" · "nobody cares about product/bottle images alone" · "nobody cares about products ranked" · "not weird enough to stop the scroll" · "brighter news room". Keeps: the NEWS-chrome styles ("Yes, more of these"; "other women in this style, and maybe Today show") and the grotesque macro nails. - 2026-05-19 — PDP DECOUPLED from
template/pdp/02_ecom("I don't like that NS is coming from the template. It should be coming from its own folder."). Never copy the template overproducts/nailscience/index.html. - 2026-05-20 — Klaviyo work: skip git (bodies live only in Klaviyo drafts).
- 2026-06-15 — RETIRED ("do not propose or do NS work").
- 2026-09-04 — the 29next store acm2 shared with Charm.
- 2026-10-13 — RELAUNCH on the new playbook (David). Base = the existing folder and site. Economics gate on the May offer: KILL at $60, passes under $35.30 — verdict pending (ask
ns-offer). - 2026-10-14 — Seven kickoff verdicts (Launch Center, Needs You): base = the existing folder and site ("yes") · checkout = Shopify ladder + 29next for trial and VIP (house default) · offer = "Rebuild the offer first, then re-run the gate" · formula = "What's on the PDP is what's in the pens" (Truth Gate claim 1 settled; the Lansil quote's salicylic line was a reference formula) · inventory = "Ignore, we have stock in Shenzhen" (drop-ship model stands, no re-quote) · accounts = "Yes same." (ad account act_832705320684189, pixel, GTM-P9MSJBP7 all kept) · guarantee = 30-Day Money-Back Guarantee everywhere ("yes"). Still open:
ns-webhook-crossfire. - 2026-10-14 (later) — OFFER LOCKED: Ladder A (David, in chat: "Ladder A is good") — $29.99 single · $59.98 for 3 kits · $89.97 for 5 kits on the BOGO formula, monthly $24.99 on 29next. The gate says FIX at $60 and passes under $43.16; David built it anyway, so $43 is the CPA target for the first rounds. Scratch candidates B and C retired.
- 2026-10-14 (later) — NO deep-research re-run (David: "Why are you doing GEMINI? The deepresearch doesnt need to be done"). The chain I had queued on the Gemini driver was stopped before any report ran and its prompts removed; the April copy brief (
tools/deep-researcher/runs/nailscience/at-ns/07_copy-brief/response.md) is the research base — Mode A. Phase 1 = the Truth Gate written from that brief + the live PDP, then the Angle Studio mined from the brief, the ad-library sweep and Reddit / Amazon (no prompt 08). - 2026-10-14 (evening) — MEDIA PLAN 2026-10 = a turn-on plan (David, with the Ads Manager screenshot: "Here are the winning ads for nailscience, i just want to run these initially on those prelanders … I dont want to launch new ads, i just want you to turn these on when we are ready. VERY IMportant to turn them on, because i dont want them dissapproved if we re-submit them" + "along with the video ad that i put in the ads folder"). The three: W1 AD031 (NY podiatrist image, Foot Doctor page, 5 purchases at $20.05) · W2 AD013 (painful nail image, Foot Doctor page, 2 at $23.07) · W3 AD024 (pharmacist's #1 pick image, Top 5 page, 17 at $24.54). HARD: never edit, duplicate or re-submit them (creative, link, copy, targeting, name) — a re-review could disapprove them; they are only switched ON, after their seven sibling ad sets are paused. All three link to thehealthreportdaily.com (the root and /blog/ranking), re-homed to the Bronze hub in July → the fix is routing, not editing: the two pages mirror to trynailscience.com/blog/foot-doctor/ + /blog/ranking/ and the old domain 301s there. The one new ad = V12 the fungus rap (
creative/nailscience/videos/video12_edited.mp4), ported copy from W3's winning pair. Bench: AD025 (the Top 5 sibling: 21 purchases at $39.43 on $828, paused in May) · AD014 (Q-Tip, 5 at $55.96). NS-P1 (5 Reasons) benched: $247 for 0 purchases in April. The Top 5 page's 60-day guarantee → 30-day in the source (Guarantee Law). Asks:ns-ads-destination·ns-video-cut·ns-ads-go. - 2026-10-14 (evening, later) — W4 added to round one: AD025, the Top 5 sibling on the fhom79 image (ad
120243267777970510, ad set120243267766830510paused since May 24; $827.94 → 21 purchases at $39.43, the most purchases of the April run). David, shown the four ads toggled on in his Ads Manager (the third was naobtu9/AD034, 4 at $28.69): "LEts do the $39.43 one actually." So round one = W1 + W2 (Foot Doctor CBO $50) · W3 + W4 (Top 5 CBO $50) · V12 ($50). The turn-on gains one toggle (W4's ad set back on). AD034 and AD014 are the bench. - 2026-10-14 (evening, last) — HOLD until David's go ("yes just wait until i give you the go ahead. For now just wait"). Nothing switches on, nothing deploys to production for this plan, until he says go; the three asks stay open on the lane.
- 2026-10-14 (night) — SANS-SERIF + BIGGER TYPE ON MOBILE (David: "On mobile, the text looks really small still. Again we want to use San Serif font for readability as its 45+ product") — the house 50+ readability law applies to NailScience: every serif rule on the PDP → DM Sans (the Clinicians' Choice card keeps its own type as a brand mark), the phone body base 14px → 17px / 1.5, one floors block on the sections' own selectors (
build_pdp_read50.py: body ≥ 17px, other lines ≥ 15px, labels ≥ 14px), the muted inks darkened to AA (--ink-3/-4→ #7a5a66), the accent used as text or a button ground → #a93f61 (the brand pink #c74e74 reads 4.4:1 on white), the comparison table fixed inside the viewport (it was the 25px sideways scroll), 44px targets on the logo, rating, Clinicians' link, footer links and the before/after dots, the hero/video/press strips clipped. Measured withtools/qa/mobile_readability.py: PDP 574 findings → 0 at 375 and 390, home/collection/legal 0. The homepage, collection and legal pages took the same accent and the sans headings through the builders. - 2026-10-14 (late) — HERO CAROUSEL = the house standard, and the urgency block is gone (David: "we need to apply the learnings of the playbook for the hero carousel, what was it?" · with a screenshot of the ⏰ 60% / 🔥 17-kits block: "remove this for all views on the pdp").
products/nailscience/build_pdp_extras.py(ported from CellScience's): the ⊕ zoom lightbox on every slide (tap opens, pinch / wheel / double-tap 2.5×, ‹ › inside, one click zooms back out), looping ‹ › arrows at 44px (the last right arrow rotates to slide 1; the template's › peek hint gone), the six 4 MB April PNG creatives stay the SOURCE files while the page reads content-hashed WebP twins (480 / 720 / 1200 / 1600 + 240 thumbs — 2.2 MB for all thirty against 25 MB of PNG; slide 1 eager + fetchpriority high + a head preload, slides 2–6 hydrate on approach), the thumb strip sorted by slide (the April strip listed 6 before 5), slides carrydata-slot07–12 for the imagery wave. The [13] urgency block AND the sticky bar's "⏰ · 🔥 left" line removed on every view (the two mid-page "Introductory offer ends in" countdowns under CTAs are still there — flagged, not cut). Measured 0 findings at 375 / 390 / 1280; staging QA at 390 + 1280: loop 1→2→3→4→5→6→1, prev from 1 → 6, slide 2 hydrates on the first arrow, lightbox opens the 1600 twin, Esc closes, no console errors. Rebuild order (§8) now: offer_box_v2 → pdp_house → wire_pdp_footer → build_pdp_extras.py → build_pdp_read50.py. - 2026-10-14 (late) — TRY BEFORE YOU BUY LANE BUILT on 29next (David: "Please do the Try Before you Buy 29next checkout flow"). The campaign answers again (the client API returned it tonight —
ns-campaign-reactivatewithdrawn). Store: the package table in §3 (Ladder A on 2/10/11 as their own products, the two trial lanes on 12/13, the $6.99 attach pending on the owner's one admin screen — askns-tbyb-shipping). Checkout (build_checkout_tbyb.py, in place over the May page): the house head block (same-origin fetch shim →/next-api/*proxy, a stale May?pkg=3|4|5rewritten onto Ladder A before anything reads it,?pkgpurges stale cart state, the foreign-cart guard, the paint-first boot), the May "Your Selection" bar and its preselect gone, ONE trial card re-pointed per lane by the house?pkgpreselect (plan=tbyb-sub|tbyb-one; the package section and the other cards hide underhtml[data-tbyb]; Shipping Protection opt-in on the trial), the radio-row selector (Buy 2 + Get 1 FREE default · Buy 3 + Get 2 FREE · 1 Kit, then "or" and the 14-Day Free Trial) withsel3-seed, the Total Order Summary under the pay button (build_order_summary.py; the bottom Item/Amount table gone), SECRET50, two missed 90-day lines → 30-Day, the footer's May.htmllinks →/pages/<slug>/. PDP (build_pdp_extras.py): the?ob=ccolumn (product head · the [07] bullets · Try it & subscribe vs Buy One · the guarantee line · the CTA to pkg 12; session-stickyns_ob), the bundle tiles → pkgs 2/10/11, the trial row → pkg 12 (offer_box.json too). Measured 0 findings at 375/390/1280 on both columns. No disclosure paragraphs on the checkout (David 2026-09-11 house call); the day-14 charge sits on the OB-C plan row. Production env set (§3). Not yet: a live trial order (needs the $6.99 attach), the Shopify ladder (askns-shopify-store). - 2026-10-14 (late) — SUBHEAD = David's line: "Two pens. One opens the nail, one kills the fungus underneath." (his pick from the options offered; two rendered lines at 375 by his call — P-02's one-line rule waived on this page). Lives in [05] and, copied at stamp time, on the ?ob=c column; COPYCHIEF row 26.
- 2026-10-14 (late) — The "No pills. No blood tests. No $800 doctor visit." benefit bullet REMOVED (David, with the screenshot: "remove this"). Three bullets remain in [07] and on the ?ob=c column; the comparison table's "No pills, no side effects" row stays. COPYCHIEF row 27.
- 2026-10-14 (late) — Offer box row sub-lines = the supply alone: "1 Month Supply" · "3 Months Supply" · "5 Months Supply" (David, with the screenshot: "change to just 3 Months Supply, or 5 Months Supply or 1 Months Supply"); the same lines on the checkout rows and the order summary (CL-10 parity).
offer_box.jsonis the words file; COPYCHIEF row 28. - 2026-10-15 — SHOPIFY STORE DELIVERED → the ladder checks out on Shopify (David: "NAilScience shopify" + the app's client id and secret). The house kit ran end to end (§3 Shopify block): the product, the shipping rules, the storefront token, the handoff LIVE on the PDP / home / collection on staging; the live matrix 16/16. The trial and VIP stay on 29next. Two owner settings open (
ns-shopify-finish): the time zone (Asia/Manila) and the Facebook / Google channels.ns-shopify-storewithdrawn (steps 1 · 2 · 3 · 5 done). - 2026-10-15 (midday) — LIVE ON trynailscience.com + THE QA STACK (David: "run full QA stack on this, and page speed optimization if needed. make sure the attribution works, capi, all of that. make sure the try before you buy lane works." · "and all of shopofy attribution works also" · "lets deploy to the main site whenever we are ready"). Production deployed 11:20 PT through
products/nailscience/deploy.sh --prod(hygiene PASS,--functions, smoke 200s/301s; the two prelander mirrors ship at/blog/ranking/and/blog/foot-doctor/) and again 12:00 with the paint-first tag boot. Proven on production: the Shopify matrix 16/16 (iPhone WebKit + desktop Chromium) · the browser attribution proof ALL PASS on the PDP / checkout / trial checkout / VIP / thank-you (verify_tracking.py browser: every event carries fbp + fbc with a matching browser/server event_id; the handoff fires AddToCart + InitiateCheckout; the Shopify cart link carries utm_* / fbclid / _fbp as cart attributes) · the webhook ALL PASS (verify_tracking.py webhook: unsigned 401 · tampered 401 · transaction.created ignored · foreign SKU ignored —ORDER_SKU_PREFIX=nailscienceends the Charm/CellScience crossfire — a signed test order 200 received, no Purchase relayed) · a dummy-card LADDER order on the 29next checkout 20/20 on iPhone WebKit (tools/qa/nailscience_tbyb_order.py --lane ladder --base https://trynailscience.com: order 104864 is_test → VIP accept → Extra Strength → Fungus Destroyer → thank-you; back office one line qty 1 $59.98 on nailscience-3kits + VIP $14.99, free shipping, total $74.97, the VIP month/1 $9.99 subscription cancelled; relay 200 received). Fixed in the round: the VIP page REBUILT from the CellScience chassis (build_oto.py— the April page was a browser "save page as" of a vendor funnel: a copied tag container, a stranger's Clarity + Klaviyo ids, dead ajax calls, 78 junk files, no shim; COPYCHIEF row 30), the runtime's Pattern D (no PageView on the receipt) + Pattern M (CAPI waits ≤ 2.5 s for _fbp/_fbc and re-reads at send), the webhook secret (the production env held a stale 20-char value while webhook 67 signs with its 12-charsecret_key; Netlify reads secrets back MASKED and a PUT keeps the old value → DELETE + POST with explicit contexts; the probe now signs with the webhook's own key), the handoff's InitiateCheckout on the link path (shared/shopify/shop-handoff.js), the checkout's one hashed CSS bundle + Swiper gone + a<picture>hero (no hero bytes on phones), the runtimedefer, the GTM + pixel loaders inserted after a double rAF (tracking_head.paint_first, every page). PSI mobile on production after the paint-first boot: checkout 38 → 83 (FCP 1.4 s · LCP 1.9 s) · home 92 (FCP 1.1 s · LCP 1.4 s) · PDP 58 (FCP 2.3 → 1.4 s · simulated LCP 9.4 → 3.5–4.6 s · TBT 1.1 s · CLS 0.2 on one run, 0.024 on the next — one "media element lacking an explicit size" in.atf-gridto size; observed LCP on a real phone 2.3 s; the hero paints ~1.9 s late behind the tag scripts' main-thread work — the next lever is GTM's own payload, not ours). The trial lane — the gateway's verdict: everything builds right (the lane card, the $0 line, the day-14 subscription) and 29next answers the $0 first order400 "Orders with subscription products require payment."→ the free trial NEEDS the $6.99 shipping attach (askns-tbyb-shipping, re-posted with the quote); the hour it exists:build_29next.py --create→build_checkout_tbyb.py→ the runner--lane sub. Found on the way: the order runners filled the Spreedly card frames withfill(), which the current frames ignore (the submit died silently, no error anywhere) — typed keystrokes fix it (NailScience + CellScience runners patched); Shipping Protection stays hidden — the May call (8ed06a76c), not a regression; the Dr. Franks mirror nested the April Top-5 twin as/blog/foot-doctor/blog/ranking/(now excluded from the sync, gone at the next deploy). Memory:29next_checkout_qa_traps.md. - 2026-10-15 (night) — David's answers in chat (to "what's left for me to do?"): #1 the $6.99 trial shipping → "You do it through my browsers." — done in his Chrome (Claude in Chrome, the 29next admin already signed in, store acm2 "NailScience Co." confirmed before creating): "Try Before You Buy Shipping", code
tbyb, $6.99 per order, everywhere, public; attached as ref 2; the free trial takes orders on production (monthly 24/24 on iPhone, order 104892: $6.99 today, the day-14 $24.99 scheduled, both renewals — the trial and the VIP — read $0 shipping). The VIP renewal can carry the $6.99 for up to an hour when 29next's order.created lands before the VIP accept (staging order 104886: $16.98) — the hourlytbyb-shipping-sweep(deployed on production, minute 17) is the designed backstop · #2 Shopify "Its done" → the time zone reads America/Los_Angeles (verified); the Facebook & Instagram and Google & YouTube channels are NOT on the store (the store's publications are Online Store · Point of Sale · Shop, and the product page + checkout send nothing to Meta or Google) → ask re-posted; neither of his Chromes is signed into this store (one is a different Shopify account without access, one is signed out — signing in stays his) · #5 the video → "thats okay" = V12 runs as cut, the 90-day line included, on the Top 5 page (his call over the 30-Day lock for this one ad) · #6 the shipping promise → "its okay" = the page keeps "ships same-day from our Ohio facility", the shipping policy keeps 7–14 business days from Shenzhen · #7 the rebuilt pages → "all good" (approved) · #4 the old-domain forwarding and #8 the webhook crossfire → he asked what they mean (explained; #8 is resolved by the live SKU gate, #4 waits on his yes) · #3 the go → not answered, nothing switches on.
- 2026-10-16 — THE APRIL PRELANDERS ARE BACK AT THEIR APRIL URLS (David, to #4: "i need the old prelanders re-deployed to the same URLs that were working before. Good catch.") — replaces the 2026-10-14 plan of 301s from the old domain to the brand domain.
tools/deploy-health-report.shnow stagesprelanders/drfranksat the root andprelanders/nailscience/top5-advertorialat/blog/ranking/of the SkinScience hub site that owns thehealthreportdaily.com (the 16 SkinScience / PeptideScience pages untouched, nothing else of theirs changed since the hub's Sept 18 deploy; draft first, then production). Source fixes: the pages' runtime sends its server events to https://trynailscience.com/api/capi when served from that host (the hub runs no functions; text/plain sendBeacon, the relay answers ACAO *; the deploy refuses a stage without it); the buttons land on the PDP/products/nail-restoration-kit/(they pointed attrynailscience.com/— the PDP in April, the homepage since the relaunch); the sticky bar re-reads the Meta cookies at click time; Pattern M ported into the prelander runtime (the server PageView left before_fbp/_fbcexisted). Proof live on desktop Chromium + iPhone WebKit (tools/qa/nailscience_prelander_proof.py, 41/41): only pixel 1341229411229569 (PageView + PrelanderView), GA4, server events to the brand relay with fbp + fbc + the browser event_id, first-party cookies, every button → the PDP with fbclid / fbp_in / fbc_in / eid_in, the same visitor on the PDP (eid adopted; the PDP's server events carry fbp_in and the ad's fbclid). The Top 5 page's buttons open the PDP in a new tab, as in April. The brand-site copies (/blog/foot-doctor/, /blog/ranking/) took the same fixes and stay live. - 2026-10-16 — THE SHOPIFY FACEBOOK & INSTAGRAM CHANNEL IS LIVE (David, after his invite to the store: "Wait im gonan setup the facebook instagram app" — he set it up himself). The page the ads run from, Carol Anne Foster, could not be picked: it is already connected to another Shopify store — the page in the Shopify app only decides where a Facebook Shop would appear, attribution rides the dataset, so he connected "Rachel's Austin - Family Journey" instead (leave Carol Anne Foster where it is). Verified: the store's channels now include Facebook & Instagram, and its checkout reports PageView + AddToCart + InitiateCheckout to dataset 1341229411229569 (the existing one). Instagram shop not connected (not needed). Google & YouTube still not installed.
- 2026-10-16 — ONE COUNT PER ACTION (the double count the channel created): the PDP's handoff and its own Add To Cart wiring ALSO sent AddToCart + InitiateCheckout for every Shopify-bound click, so Meta would have counted both twice — and the media plan judges ads on cost per checkout started ≤ $4. Now
creative/nailscience/shopify-spec.jsonhandoff.store_reports: ["meta"]→shared/shopify/shop-handoff.js(house, v2:store_reports,ShopHandoff.bound()/.report()) sends a Shopify-bound AddToCart / InitiateCheckout to GA4 only (dataLayer push, no Meta, no CAPI); the trial row (29next) still reports to both. Found on the way: the PDP's AddToCart still carried the May ladder ($21.99 / $39.98 / $53.97 on packages 2/3/4) → Ladder A values and packages 2/10/11 (+ the trial row pkg 12 at $0) viabuild_pdp_extras.pyatc_wiring.verify_tracking.py browsernow checks both halves — the PDP sends Meta nothing for a Shopify-bound click and GA4 both events at $59.98; the Shopify checkout reports to dataset 1341229411229569 — ALL PASS on staging and production. Add"ga4"tostore_reportsthe day the Google & YouTube channel is on. - 2026-10-17 — THE BUNDLE ROWS TAKE THE APRIL NAMES (David, after asking what the bundles were called before the relaunch — "Mild Issue / Moderate Issue / Full Treatment" on the May ladder: "How do we change the bundle lines to be like that?"). The row titles only — Ladder A's prices, badges (Most Popular / Best Value), savings and the supply sub-lines (his 2026-10-14 call) stay: 1 Kit → Mild Issue · Buy 2 + Get 1 FREE (3 kits, default) → Moderate Issue · Buy 3 + Get 2 FREE (5 kits) → Full Treatment. CL-10 parity everywhere a shopper meets them: the PDP box (
offer_box.json→offer_box_v2.py, then the documented chain —pdp_house.pywith--accent 'var(--accent, #c74e74)'), the cart drawer's first paint (build_pdp_read50.py), the homepage tier cards (build_site_pages.py, kit counts kept in the sub), the 29next checkout rows (build_checkout_tbyb.py), the 29next packages renamed IN PLACE (ids 2 / 10 / 11 kept;build_29next.pyRENAME_FROM; the order summary reads them: "NailScience — Moderate Issue (3 kits)"), and the Shopify variants renamed in place byshopify_store.py --align("Moderate Issue (3 Kits)"; variant ids, prices and SKUs unchanged). Readability 0 on PDP / home / checkout; live on staging + production 2026-10-18; checked live on every surface. The warehouse product titles keep the kit / pen counts. COPYCHIEF row 31. - 2026-10-18 — NO PRICE ON THE ADD TO CART BUTTON (David, with the screenshot "ADD TO CART — $59.98": "remove the price from add to cart" — the house rule since 2026-09-28, which NailScience never got because it was dormant).
build_pdp_extras.pyatc_labels: the static label and applyBundle()'s per-row rewrite both say "Add To Cart" (the trial row keeps "Try It for Free"); the same step finishes the April names in the sticky bar and cart drawer titles (they reverted to "3 Kits (Buy 2 + Get 1 FREE)" on a row click). Checked live on desktop + iPhone, every row; readability 0. COPYCHIEF row 32. - 2026-10-20 — THE SINGLE KIT IS $24.99 (David, with the Mild Issue row screenshot: "set the base 1 unit price to 24.99"). The single $29.99 → $24.99 against the same $49.99 compare-at (Save $25, 50%); the bundles stay at $59.98 / $89.97 (not cascaded — 3 kits is now 2.4× the single, not the BOGO 2×; his call if he wants $49.98 / $74.97), the monthly lane stays $24.99. Every surface from one source —
build_29next.pyUNIT_PRICE → the map'sladder_prices→build_pdp_extras.pyprices()(tile totals, the applyBundle table, the ATC values;offer_box_v2.pyre-renders the per-kit price + Save pill from the tile totals, so the PDP chain runs twice after a price change),build_checkout_tbyb.pyrows (from the map),build_site_pages.pyBAG_PRICE + the home / collection meta,pages_content.pyONE_PRICE (Terms §3.2's day-14 charge) +check_compliance.py,build_oto.py(the VIP page's kit value $24.99 / Save $10.00 — the $14.99 VIP price kept), the Shopify spec →--align(variant ids kept),build_29next.py --align(pkg 2 $24.99; pkg 13's day-14 keep $24.99 withis_recurringsent explicitly — the align PATCH now always sends it, the Sculptape 09-18 trap). Knock-ons worth his eye: (1) the OB-C subscribe plan's "Subscriber price of $24.99 (vs $29.99)" lost its contrast — the row now reads "Locked at $24.99 a kit for as long as you stay" (the monthly lane no longer undercuts the single; drop the monthly to restore a gap, or leave it); (2) the VIP "one kit at 50% off" formula would be $12.49 on a $24.99 kit — kept at $14.99 (40% off), unasked; (3) the April prelanders still say "Only $29.99" (Foot Doctor) and "single kit is $21.99 … enough for 6 months" (Top 5) — ads live, untouched, his open call from 10-16. Economics gate re-run: FIX at $60, passes under $43.16 (unchanged — the monthly lane drives it); the average first order now leaves $39.27 (was $40.80). Proven on staging and production: PDP rows / button / drawer / ?ob=c column / home cards / 29next rows / the one-kit trial card / the Shopify tier-1 checkout all $24.99, no $29.99 left; a dummy one-kit trial order books with the $24.99 day-14 charge (pkg 13 recurring intact); readability 0 on PDP / home / checkout. Mid-round: round one has been live since 10-16 22:41 PT — day 4 of the day rules straddles the price change. COPYCHIEF row 33. - 2026-10-21 — THE LADDER CHECKS OUT ON 29NEXT AGAIN; THE SHOPIFY HANDOFF IS DORMANT (David, 2026-10-21: "lets make the checkout like it was on 29next, not shopify" — reverses his 2026-10-14 checkout verdict for NailScience; the house default (Shopify ladder) stands for the other brands). One switch:
handoff.on: falseincreative/nailscience/shopify-spec.json→tools/shop_handoff.pyre-stamps the PDP / home / collection DORMANT — the ladder's Add To Cart follows the page's own link into/checkout/?pkg=2|10|11(the house 29next checkout: Ladder A rows, the Total Order Summary, the trial card, SECRET50), no Buy with Shop Pay under the button, the?ob=cBuy One → pkg 2;?co=shopifypreviews the Shopify path. Attribution: the PDP's own Add To Cart wiring reports through the runtime again (Meta + GA4 + CAPI — thestore_reportsrule only applied to Shopify-bound clicks), the 29next checkout fires InitiateCheckout on load, the Purchase comes from the order webhook (the SKU gate) as on every 29next brand; Shopify's channel reports nothing now (no Shopify orders). Proofs:verify_tracking.py browser(a dormant branch: AddToCart with fbp + fbc + event_id, one navigation into /checkout/?pkg=10 carrying utm_* + fbclid, no Shopify request, no Shop Pay), the row route check on iPhone + desktop (every row → its package, the SDK's total),nailscience_tbyb_order.py --lane single(pkg 2 at $24.99 books on 29next → VIP → OTOs → thank-you → relay) and--lane ladder, the matrix on the?co=shopifypreview (qa.query). The store, its product and the Facebook & Instagram channel stay as they are (flip back = removeon+ re-stamp). Readability 0. - 2026-10-21 — THE CHECKOUT NEVER ASKS FOR THE PACKAGE AGAIN (David, 2026-10-21, with the checkout's "Choose Your Package" screenshot: "If the package has been selected, then on checkout we should not be having them select the package again. That doesn't make any sense."). A
?pkg=link that names a ladder row now collapses the selector to that one row under a "Your Package · Change" header — the other rows, the "or" divider, the sub-line and the trial card hide; Change brings the full selector back (the SDK is never touched, so a re-pick after Change swaps the cart as before); a bare/checkout/, a rejected id and the trial lanes keep their shapes (the trial section was already hidden, the Total Order Summary carries its line). Built once, in the house moduletools/checkout_selector.py(the handoff state:sel3-pickin the head setshtml[data-sel3-pick]before the first paint, static per-row CSS shows only that row,sel3-pick-selright after the cards makes sure that row is the selected one before any SDK boot) — default on wherever the section is titled "Choose Your Package": NailScience now; CellScience, Charm and PeptideScience on their next rebuild (dry-run clean); Sculptape's cart-mode "Your Cart" stays as it is unless its spec opts in. The same pass fixed the selected row wrapping on desktop ("Moderate / Issue", "Free / Shipping" in his screenshot): the module's no-image grid class matchedclass="os-card bundle-cardliterally and the default card readsos-card next-selected bundle-cardafter the first build, so every rebuild since 10-14 had dropped it. Standing checkout law from here ([[feedback_checkout_no_repick]]). Proofs: the collapsed-selector check (A–F: pkg 10 / 2 / a May id / bare / the trial lane / a bad id, then Change → row 11 swaps to $89.97) on iPhone WebKit + desktop Chromium, the attribution proof in the browser and on the webhook, the four order lanes, readability 0 — staging, then production, all passed (the folders are in §10's 10-21 line). - 2026-10-21 — THE CLINICIANS' CHOICE CARD IS OFF THE PDP; THE SMALLER BADGE IS ON THE BOARD (David, 2026-10-21, with the card's screenshot: "remove this lets figure out a smaller badge that we can use"). The [05b] card (the serif brand mark he kept on 10-02) left the product head of the PDP source, its CSS and the read50 / pagespeed rules with it; the claim it carried (658 clinicians, including board-certified dermatologists, recommend NailScience to their patients — without compensation) is NOT dropped: five smaller forms are on the board
/docs/nailscience/clinicians-badge/board.htmlon the Launch Center — A a trust line under the rating (full sentence, 44px tap) · B a 44px pill on the paper tone · C a 44px gold-ring seal with the serif mark · D a 48px avatar chip with the three faces · E the count folded into the rating line (no added height) — each rendered in the page's type at 390 and 560 with height, smallest text and contrast measured live; askns-clinicians-badge. Until his pick nothing replaces the card; the template (template/pdp/02_ecom) keeps its full card for the other brands. Readability 0 at 375/390/1280 without it. - 2026-10-21 (afternoon) — THE OFFER BOX ROWS: NO STRUCK PRICE, NO FREE SHIPPING LINE, NO TRIAL ROW; THE CLINICIANS CHIP (BOARD D) IS IN (David, three screenshots of the PDP offer box: "remove the strike-out price, remove the Free Shipping," · the 14-Day Free Trial row: "remove this section also" · lane D of the badge board: "i like that one to be reintroduced across this page and the v1 pdp of cellscience"). Rows: each ladder row reads title · supply · per-kit price · Save pill — no struck $49.99, no truck line (
products/nailscience/offer_box.json: noshipon the tiers,strike: false;tools/offer_box_v2.pyradio layout gained the switches, defaults unchanged elsewhere). Trial row: off the standard box (trial_row: false— the config stays for the?ob=ccolumn, the trial links andplan=tbyb-*; the trial lanes, packages 12/13 and the $6.99 attach are untouched, the trial is simply not offered inside the standard box). The checkout rows mirror all three (CL-10;tools/checkout_selector.pyflagsstrike/ship_line/trial_pick, set inbuild_checkout_tbyb.py): no struck anchor, no Free Shipping line, the trial card and its "or" stay in the DOM for the trial lanes but never show in the full selector. The chip: the board's lane D under the rating of the product head — three faces (av-cris / av-megan / av-rosemary) · "658 clinicians recommend NailScience" · › → #reviews; 48px, 16px type, ink-2 on white, one tap target; source markup[05b] CLINICIANS CHIP+ CSS in the PDP (no builder writes it). "Across this page" is taken as the one placement the board showed; a second placement near Add to Cart is a one-liner if he wants it. Askns-clinicians-badgeclosed by his chat verdict. Readability 0 at 375/390/1280 on the PDP and the checkout. Proofs: staging — pick_check A–F on iPhone WebKit + desktop Chromium ALL PASS (three rows after Change, no struck price / no Free Shipping visible, the trial lane unchanged), the row route check ALL PASS (rows → /checkout/?pkg=10|2|11 at $59.98 / $24.99 / $89.97, no Shopify request), attribution in the browser ALL PASS, ladder order 20/20 (ns-tbyb-20261021-163652); production — pick_check ALL PASS (screenshots + the PDP head and offer box shots innailscience-pick-20261021-pm), the route check ALL PASS, attribution in the browser ALL PASS - 2026-10-22 — THE NAILSCIENCE CHECKOUT: NO PACKAGE SECTION ON A PRODUCT-PAGE ARRIVAL, NO STRUCK RETAIL, NO ORDER SUMMARY (David, 2026-10-22, NailScience only — "THese are anilscience checkout updates"). (1) On the one-row "Your Package · Change" state and the summary bar's "$59.98 / ~~$149.97~~": "this is still showing and is still showing, how come these arent gone?" — his 10-21 sentence ("we should not be having them select the package again") meant GONE, not collapsed. A
?pkg=row arrival now shows no package section at all (tools/checkout_selector.pyhandoffmode: "hide": the section holding the cart selector carriessel3-section,html[data-sel3-pick]hides it, the selection flip still runs so the SDK seeds the linked package); a bare/checkout/keeps the three-row selector. The module default stayscollapseso no other brand's tested behaviour moves without its owner's word — new adopters takehide. The struck retail left the summary (build_order_summary.pySTRIKE = False). Proofs: staging pick_check A–F ALL PASS (iPhone WebKit + desktop Chromium: no package section on pkg 10 / 2 / a May id, bare /checkout/ keeps three rows and a re-pick swaps to $89.97, the trial lane unchanged, no struck number visible), route check ALL PASS, attribution browser ALL PASS, ladder order 20/20 (ns-tbyb-20261022-072244); production pick_check + route check + attribution ALL PASS (29 checks, screenshotsnailscience-pick-20261022). (2) Then, on the bar itself ("TOTAL ORDER SUMMARY ▾ $89.97"): "remove this" — the Total Order Summary is OFF the NailScience checkout (build_order_summary.strip_text()frombuild_checkout_tbyb.py: component, CSS, JS; the bottom Item/Amount table stays gone;build_checkout_read50.pyno longer requires the summary CSS). The trade-off, flagged to him the same turn: with the package section hidden on a product-page arrival and the summary gone, nothing on the checkout names the package or states the total before Complete Purchase (the trial lane's $6.99 today included) — the PDP's row and the 29next receipt are the only places the shopper reads it; one line under the pay button would restore it if he wants. The house structure ([[checkout-total-order-summary-structure]]) stands for the other brands. QA tools now read the SDK's own cart (next.getCartData().cartTotals,sessionStorage next-cart-state) —tools/qa/nailscience_tbyb_order.pywait_sdk+ checks. Readability 0 at 375/390/1280 on/checkout/?pkg=10and the trial lane. Proofs: staging pick_check ALL PASS (no package section, no summary on the page, the cart line = the package; bare /checkout/ three rows; the trial lane $6.99), route check ALL PASS, attribution browser ALL PASS, ladder order 20/20 (ns-tbyb-20261022-072953), trial monthly booked 25/26 twice (ns-tbyb-20261022-073033,-073205) — the one miss is the renewal-shipping read on STAGING (29next's delivery had not carried the subscriptions yet; the runner's own replay hits the staging relay, which has no store key); a signed replay of the same order to the production relay zeroed it to $24.99 at once, the production relay verifies the store's current webhook key, and the hourly sweep finds 0 live trial subscriptions carrying shipping — not a product fault, unrelated to the summary; production pick_check + route check + attribution ALL PASS (29 checks, screenshotsnailscience-pick-20261022-notos) - 2026-10-22 — ATTRIBUTION VALUES FIXED: THE CHECKOUT FUNNEL EVENTS AND VIEWCONTENT CARRIED THE MAY $29.95 (David, 2026-10-22: "can you make sure attribution is working correctly also"). Measured on production before any change — the pixel and the CAPI relay recorded, nothing sent: every InitiateCheckout, shipping_info_filled and AddPaymentInfo carried value $29.95 whatever the package (the May checkout's
getSelectedPackage()read a.activecard with a[data-next-price]price — the house selector has neither — so the|| 29.95fallback fired every time; it predates this week), and the PDP's ViewContent sent a hard-coded $29.95. AddToCart was already right ($24.99 / $59.98 / $89.97), and the Purchase uses the order's own total withpurchase_<order number>on both the browser and the webhook (dedup key verified). Fix, in the builders:build_checkout_tbyb.pystep 8ck_values()— the package and value come from the SDK's own cart (next.getCartData().cartTotals,sessionStorage next-cart-state), InitiateCheckout waits for the cart (guarded 250 ms poll, max ~6 s), the map's totals per package are the fallback (NS_PKG_VALUE, re-rendered from29next-map.jsonon every run), gate"29.95" not inthe page;build_pdp_extras.prices()— ViewContent = the single-kit price from the map ($24.99, the house convention: CellScience sends its one-bag price) and the comparison table's "Cost per treatment" cell, which the 10-20 price change had missed (customer-facing $29.95 → $24.99). The trial lane's events carry $6.99 (what the cart charges today). Found on the way: Meta's pixel stays silent under an automated browser (navigator.webdriver) — the probes hide the flag so the pixel side is measured; and PeptideScience's checkout carries the same lookup (|| 29.95) — flagged as a separate task, not touched. Proofs: staging + production (57 checks,tools/qa/results/nailscience-attribution-20261022/): pixel value == CAPI value == the cart total with event_id parity and fbp + fbc on ViewContent ($24.99), AddToCart ×3 and InitiateCheckout / shipping / AddPaymentInfo for pkg 2 / 10 / 11 and the trial lane ($6.99), desktop Chromium + iPhone WebKit; the house browser proof ALL PASS; the webhook relay ALL PASS; the prelander proof 41/41 (ns-prelanders-20261022-075213); every row's route into the checkout carries utm_* + fbclid - 2026-10-22 — META PURCHASES SINCE OCT 20 INCLUDE MY QA DUMMY ORDERS: 17 IN ADS MANAGER, 6 REAL (David: "are we double firing conversions? or did you fire them and then fired again somehow?"). Read-only audit (pixel
/stats, ad-account insights, 29next + Shopify orders): real orders are NOT double counted — each sends one browser + two or three server Purchases sharingpurchase_<order number>, which Meta merges into one. The QA dummy orders ARE counted: everytools/qa/nailscience_tbyb_order.pyrun (staging or live) sent a browser Purchase (the runner blocks no tracking; WebKit isn't flagged as automated) plus a server Purchase from the LIVE order webhook — its test-order gate readsorder.is_test, which the QA replay carries (admin-API shape) but 29next's real order.created delivery evidently does not (staging test orders produced server Purchases; staging's relay has no Meta key). Per day — Oct 20: real 1 (Shopify #1001, the Top 5 prelander) + QA 1 → Ads Manager 1 · Oct 21: real 2 (#104971, #104987) + QA 8 → 9 · Oct 22 to ~09:30: real 3 (#105018, #105020, #105034) + QA 4 → 7. Spend Oct 20–22 $489.41: Ads Manager cost per purchase $28.79, real $81.57. Same thin payload explains the 8c miss (no subscriptions on delivery) and leaves the SKU brand gate blind (it readssku; real lines carrystockrecord.partner_sku) — a latent cross-brand leak; no other brand sold on acm2 since 10-14. No fix applied yet (his question, not an ask). No more dummy orders on staging or live until the runner blocks Meta + our relay and the webhook confirms the test flag on the full order. Events already in Meta cannot be deleted. - 2026-10-24 — THE PDP BANNER SAYS BLACK FRIDAY (David: "replace \"LAUNCH WEEK\" => \"BLACK FRIDAY\" on the pdp"). The product page's announcement bar now reads "BLACK FRIDAY · DISCOUNT: SECRET50 · ENDS IN <timer>" — the bar lives in the PDP source itself (no builder writes it;
build_pdp_read50.pyonly sizes it), edited in place; the chain--checks stay current; readability 0 at 375/390/1280; live on trynailscience.com (two lines on a phone, as before). The homepage and/collection/still say LAUNCH WEEK — their bar issite_chrome.pyBANNER, left as it was (his word was the PDP). - 2026-10-25 — THE ROWS ARE MILD / MODERATE / SEVERE FUNGUS (David, with the offer box screenshot: "Moderate Fungus, Severe Fungus, Mild Fungus"). The row titles only — prices, badges, savings and the supply sub-lines stay: Mild Issue → Mild Fungus (1 kit) · Moderate Issue → Moderate Fungus (3 kits, default) · Full Treatment → Severe Fungus (5 kits). Same parity scope as the April names: the PDP box (
offer_box.json), the sticky bar + cart drawer titles on every row click (build_pdp_extras.atc_labels— any earlier name maps to the current one), the drawer's first paint (build_pdp_read50.py), the homepage cards (build_site_pages.py), the 29next checkout rows (build_checkout_tbyb.py), the 29next packages 2 / 10 / 11 renamed IN PLACE (build_29next.pyLADDER + RENAME_FROM →--align; the cart lines, orders and receipts read "NailScience — Moderate Fungus (3 kits)"), the Shopify variants renamed in place (shopify-spec.json→shopify_store.py --align; ids, prices, SKUs unchanged; the store stays dormant). Every title on one line at 390 and 1280; readability 0 on PDP / home / checkout; compliance gate PASS; checked live on every surface (names, sticky, drawer, homepage, bare checkout rows, cart lines per package) on iPhone + desktop, plus the checkout states, the row routes and the attribution proof — staging and production. No dummy orders (the QA-order leak is still open). COPYCHIEF row 34. - 2026-10-26 — ATTRIBUTION INVESTIGATION: THE FUNNEL PASSES EVERYTHING; THE LOSS IS BUYERS WHO SWITCH BROWSER (David: "lots of orders came in, but no attribution being passed back, can you investigate"). Read-only. Since 10-22: 16 NailScience orders on 29next, 2 of them cancelled + refunded on 10-23 (#105117 $59.98, #105119 $24.99) → 14 paid. Meta gets a Purchase for every order (the webhook server-side for all; the browser for most) — nothing is dropped in transit. 29next attribution: 9 of 14 paid orders carry the ad's UTMs + fbclid; all 9 were bought inside the Facebook / Instagram in-app browser minutes after the ad click. The other 5 (and 5 of the 7 since 10-24) reached
/checkout/?pkg=Nwith no ad parameters: mobile Safari or desktop, the browser's first visit to the site minutes before the order, no click ID — buyers who met the ad in the app and came to the site on their own in another browser — plus one desktop buyer back 1.5 days after his click (the click cookie survived, so Meta still credited him). The funnel itself passes everything:tools/qa/prelander_forward_check.pyon the live Top 5 and Foot Doctor pages — 9/9 buttons carry utm_* + fbclid + ids to the PDP, 6/6 PDP checkout links carry them on. All active ads point at thehealthreportdaily.com/blog/ranking. Meta credit: Ads Manager — 10-23: 1 of 1 paid · 10-24: 2 of 2 · 10-25: 0 of 1 · 10-26 (to ~13:30): 2 of 4 (server matches can land hours later). Every order WITH a click ID was credited; of the browser-switchers, Meta matched about 1 in 4 on the buyer's details. Matching keys on Purchases since 10-22: email 52/52 · first + last name 48 · ZIP 48 · city 32 · state 22 · country 16 · external_id 52 · phone 0 (the checkout almost never collects one: 1 of 16 orders). The 29next order holds IP + device + the Meta browser cookie on every order; whether 29next's order.created DELIVERY carries them to the relay is unverified (29next exposes no delivery log). Levers (his call, none applied): (1) phone at checkout — the biggest matching key missing; (2) carry the stored first-touch ad parameters into the checkout link when a later visit has none (restores 29next attribution for same-browser returns); (3) the webhook reads the full order from 29next before sending — IP, device, cookie, click ID guaranteed, and it also closes the QA-order leak and the trial renewal step (10-22). - 2026-10-26 — ATTRIBUTION FIXES LIVE: THE LAST AD TOUCH REACHES THE CHECKOUT; THE WEBHOOK READS THE FULL ORDER (David: "do 2 and 3"). (2) The last ad touch: the site runtime (
products/nailscience/tracking/ns-attribution.js) stores every ad key a visit arrives with (utm_* · fbclid · gclid · ttclid · msclkid · campaign / adset / ad ids) as_ns_adq(cookie + localStorage, 30 days, overwritten by the next visit that brings keys); the checkout's head shim (build_checkout_tbyb.pyBOOT, gated to run before the SDK boot) puts them back on a checkout URL that has NONE, before the 29next SDK reads it — one place covers every route in (PDP buttons, sticky, drawer, the homepage path, a direct /checkout/). Proof (tools/qa/nailscience_adtouch_check.py, iPhone WebKit + desktop Chromium, staging + production): a return visit and the homepage → PDP path reach the checkout carrying the stored utm_* + fbclid, a fresh browser gets none invented, a visit's own new keys win; 29next's ownnext-attributionstate shows the restored campaign / ad / ad set. Same-browser returns only — a buyer who switches browser still arrives empty (Meta matching is the only bridge; phone at checkout is the open lever). (3) The webhook reads the full order:shared/functions/order-webhook.mjsstep 3.55, opt-inNEXT_FETCH_FULL_ORDER=1(set on NailScience prod + staging viaset_prod_env.py; every other brand's behaviour unchanged until it opts in): fetch the order by number from the admin API (4 s timeout; a trial-lane order with no subscriptions yet is fetched once more after 1.5 s), its fields win; the SKU gate readspartner_skuwhen on; any failure keeps the delivery. Proofs:tools/qa/order_webhook_fullorder_test.mjs7/7 (a real order relays one Purchase WITH IP + device + fbp + fbc from the full order · a test order relays nothing and 8c still zeroes the trial sub · a foreign SKU is ignored · admin API down → delivery kept, Purchase relayed · flag off → unchanged); production webhook gates ALL PASS; two staging trial test orders through the LIVE relay (#105414, #105417) — zero Purchases in Meta for either (the test gate now holds on the real delivery) and the trial + VIP renewal shipping zeroed from the fetched subscriptions (26/26). The QA order runner now aborts Meta / GA and answers our relay locally, and its renewal check polls up to 120 s (the delivery trails the order). Dummy orders are allowed again — the 10-22 guard is lifted (both halves fixed and proven). - 2026-10-29 (15:00 PT) — Truth Gate + Angle Studio r1 built, both on the hub for David's verdicts; the statics round waits on his GO angles (David: "lets run the statics image generation playbook for NailScience and link to the best prelanders nad headlines in a media plan next"). The wild-statics brief found no Truth Gate, no angle bank, no compliance files and no product anchors (CL-20 / CL-01 stop a slate without them), so Phase 1 ran today in Mode A (the April copy brief + the live pages, no deep-research re-run per 2026-10-14): Truth Gate
briefs/nailscience/truth-gate.md+truth-gate-claims.json— 53 claims in seven groups (problems · rivals · the noun and the cause · ingredient studies · benefits and timelines · proof and persons on the live pages · offer), ⚠ on every number Claude could not source (the 200,000+ customers, 658 clinicians, two different "as featured in" lists, the 3,214 Trustpilot reviews, three fictional doctors across one funnel, the unverified FDA letter, "the only kit", "not a cure" vs "kills the fungus", which price ladder is live, SECRET50); askns-truth-gate; pagedocs/nailscience-truth-gate.html; the brand'ssources.truth_gateset in brands.json. Angle bankbriefs/nailscience/angles/— the noun the Keratin Shield (the PDP's own frame); 12 seeds (A01 the shield · A02 the disclaimer on the back of the pen · A03 the nail-softening step your cream skipped · A04 week two · A05 closed-toe heels at a beach wedding · A06 the drugstore softener stops halfway · A07 no pills, no blood tests · A08 the shoes never got treated · A09 the sock test · A10 the sandal-season head start · A11 the Army's foot medicine · A12 clipping her mother's nails), evidence = the nail-fungus ad sweep (30 creatives), the April swipe keeps, the account's results, the 53 claims (VoC: the PullPush Reddit pull did not return in time, so the problem words are the brief's; re-distil when voc/ fills); the hook-bucket brief (8 buckets × 5 hooks + the weird twenty,buckets/_brief.json, check clean); 60 batch-1 hooks (five devices, three-plus buckets, one mechanism row per angle) through both chiefs (COPYCHIEF.md: 56 PASS, 4 cut to 12 words in place, 0 killed) and batch 2 (34 rows: the three rewrites + the weird twenty on their angles + eight added rows so every batch spans three buckets,chief/batch-2.json); boardsnailscience--angles(round 1),--hooks(94),--buckets,--copysheet(122 rows); askns-angle-board-r1. The read since the relaunch (Oct 16–29, act_832705320684189): $2,342 · 29 buys; NS-P2 the Top-5 ranking is the only page with buys (27 on $1,807, CTR 6.4%, LPV $1.88, CPA $66.9, ROAS 0.74); the PDP direct 2 on $330; NS-P3 Dr. Franks 0 on $205. The headline "Pharmacist's #1 Nail Kit of 2026" carries 27 of 29 buys; Meta's asset split gives "NY Doctor: Why Your Nail Cream Is Failing" 8 buys on $71 (CPA $8.8, a quirk to re-read with volume); the body "The yellowing, the thickness, the self-consciousness… it's not your fault the creams…" carries 28. Best ads: the animated "nailcasm V2" (5 buys on $121, ROAS 1.84), the AI UGC "Solo pod" (4 on $178), the static fhom79… (12 on $491). Next, in order: David's GO angles →wild.pyGate 1 (buying situations) → Gate 2 (metaphors + formats) → the slate → render (tools/angle-statics-gen.py) → the J&T board → the media plan round that links NS-P2 with those headlines. Run file started:briefs/nailscience/statics/wild/(dest NS-P2; angles A01 / A03 / A05 as candidates until GO).
- 2026-10-29 — Angle NS-A01 It wasn't the medicine. It was the shield.: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A02 The disclaimer on the back of the pen: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A03 The nail-softening step your cream skipped: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A04 Week two on a toenail treatment: what you should see: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A05 Closed-toe heels at a beach wedding: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A06 The drugstore nail softener stops halfway: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A07 No pills, no blood tests: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A08 The toenail fungus came back because the shoes never got treated: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A09 If your toenail catches on your sock, it isn't age: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A10 Your toenail needs a six-month head start on sandal season: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A11 The Army's foot medicine: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 — Angle NS-A12 Clipping her mother's nails: GO (David, in chat ~15:15 PT: "Go on all of those angles lets see"; logged on the board by Claude)
- 2026-10-29 (17:00 PT) — Wild statics r1 RENDERED and on the hub for keeps; media plan round two proposed on the keeps (David: "render all of them lets see", after GO on all twelve angles and "Go on all of those" at Gate 1). Run
briefs/nailscience/statics/wild/2026-10-29-a.json: Gate 1 eight buying situations (real Reddit quotes), Gate 2 thirty metaphors + thirty-one formats (sixteen saved frameworks, fifteen new), the slate ten concepts + the weird twenty (every line chiefed,wild.py check0 fail) — all kept on his word;register→ hooks.md batch 3 (30 lines, ledger from the slate) →shots→tools/angle-statics-gen.py(gemini-3-pro-image-preview, product anchors cut from the supplier shots intocreative/nailscience/identity/images/pack-kit + pen-in-hand) → 30 cards increative/nailscience-ads/r1/→ the QA sheet read at full size: four re-renders before boarding (r1-16 / r1-19 / r1-20 / r1-26 leaked the prompt's own labels or instruction text into the scene; r1-30 doubled the sub; r1-26 again for a shoe-brand logo) →wild.py board(J&T manifestcreative/nailscience-ads/manifest.json, round 1, campaign NS-P2) →statics_board.py build→ https://launch-center-hq.netlify.app/docs/nailscience/boards/statics-r1/ (W3 as the anchor; need 7); askns-statics-r1-judge. Media plan round two (twin + doc): the keeps run to NS-P2 on the proven pair — headline "Pharmacist's #1 Nail Kit of 2026" + the "yellowing, thickness, self-consciousness" body, ported verbatim — seven to an ad set at $50 (his Oct 16 rule), created paused on his go, on at midnight; his calls: the keeps · the copy (the NY Doctor headline as a second?) · the open third $50 of round one or a new campaign · the go. brands.json: a second J&T boardnailscience-ads(creative/nailscience-ads). The Truth Gate stays open (ns-truth-gate).
- 2026-10-29 (17:15 PT) — Wild statics r1 JUDGED and the keeps LAUNCHED (David, in chat: "okay judged the kill, rest launch for nailscience for $50/day with the best headlines and prelander"). The board
nailscience--statics-r1: 15 kills — r1-02 pharmacy receipt · 03 wedding Polaroid · 04 pill / bottle / pens panel · 07 kitchen calendar · 08 paint chips · 09 race bib · 13 seed packet · 14 blood tube + insurance letter · 16 hourglass curiosity post · 21 failed-creams counter · 22 four-panel grid · 24 three-doors article · 27 hardware aisle · 29 sandals-with-tags portrait · 30 1970s primer ad; the 15 he did not touch = the launch set. The taste read: the loud saved frameworks held — anatomy callout, wikiHow, the surgical engraving, the question post, handwriting on skin ×2, the news desk, the rotting organ, the split ×2 (10 of 16 saved-framework cards launched) — while every quiet photo framework died (confession portrait, failed-remedies still life, quad grid, medical article, curiosity bait, expert split: 0 of 6) and the NEW paper-object formats mostly died (receipt, calendar, paint chips, seed packet, race bib, Polaroid, letter: 5 of 14 launched; the survivors — the 1970s shield ad, the door-frame growth chart, the care label, the pedicure menu, the locker-room notice — are now saved frameworks increative/dashboard/image-frameworks.json). By situation: "It came back" (NS-A08: the shoes, the shower floor, the bath mat, the boomerang, the greenhouse shoe) 5 of 5; "The thick nail" (NS-A09) 3 of 5; "Nothing has worked" (NS-A01) 2 of 4 — the physical barrier (cast, shield) lives, the product counters die; "The podiatrist's last option" (NS-A03) 2 of 4 — the clinic's order lives, the primer metaphor died twice; "The slow clock" (NS-A04) 2 of 4 — growth visuals live, the 21-day / week-two promises die; "Hiding the feet" (NS-A05) 1 of 4; "The pill and the bottle" (NS-A07) 0 of 3. The launch: campaign120250100703970510"NailScience — NS-P2 Top 5 · wild statics R1 keeps · CBO $50/day", two ad sets by what the card says — "What it really is" (8: where the fungus came from + the thick nail; ad set120250100705080510, AD036–AD043) and "Why the cream failed" (7: the barrier, the clinic's order, the growth clock, the polish; ad set120250100706820510, AD044–AD050); every ad on the proven pair verbatim (headline "Pharmacist's #1 Nail Kit of 2026" + the yellowing body, read from AD024/AD025's asset feed at launch); the page = the Top 5 at its April URL (the manifest's live_url); 18–65 US, Advantage+ audience off, FB / IG / Messenger; created PAUSED 17:12 PT, ACTIVE at 00:00 Oct 30 PT (agentcom.davidsanghera.ads-on-20261030-0000, with MGI's retargeting and WAM's Joe ads). His "$50/day" is the campaign (the open third $50 of the $150 cap) and fifteen does not split into sevens, so the round went CBO with the ad sets by story (the CellScience r7 precedent) rather than two ABO sets at $50 each; "the best headlines" = the one headline that carries 27 of 29 buys — the NY Doctor headline stays a later round's variable (one variable a round). Planbriefs/nailscience/launch/2026-10/NS-P2-STATICS-R1-CBO.json+.state.json; copy sheet rebuilt (the live rows). - 2026-10-29 (17:20 PT) — Round two of the weird statics ordered, on the saved frameworks (David: "and lets generate some more weird ads, using our frameworks also"). Run
briefs/nailscience/statics/wild/2026-10-29-b.json, built from the r1 taste read: weird metaphors two rungs past, every card on a loud saved framework (the eight that held plus the three unused — the painted-disgust illustration, the 3D villain, spots-spell-a-word — and the five NEW formats that survived r1), weighted to "It came back", "The thick nail", "Nothing has worked", "The clinic's last option" and "The slow clock", plus the places the fungus is caught (the salon footbath, the shared shower) and the drugstore softener that stops halfway (NS-A06); no pills-vs-pens, no hiding-the-feet portraits, no paper-object counters. Rendered whole on his standing "render all of them" and boarded as statics r2 for keeps. - 2026-10-29 (21:48 PT) — Weird statics r2 JUDGED and the keeps LAUNCHED as round three (David, in chat: "i kept some, we can launch a $20 ad for those next" → "starting midnight tonight"). The board
nailscience--statics-r2: 7 keeps, nothing killed, 21 untouched — r2-06 the shower-drain news desk · r2-08 the hotel-carpet rotting toenail · r2-10 the hoof engraving · r2-11 the barnacle anatomy callout · r2-13 the sealed-jar handwriting · r2-15 the locksmith news desk · r2-23 the half-shoveled-driveway split. The read: the news desk kept 2 of 2, the rotting organ 1 of 1, the engraving 1 of 2, the anatomy callout 1 of 3, handwriting 1 of 2, the split 1 of 2; nothing from wikiHow (0 of 4), the 3D villain (0 of 3), the painted-disgust plate, the spelled nail, the 1970s ads, the locker notice, the salon menu, the question post, or the two NEW looks (the wanted poster, the museum vitrine — not saved). By story: "It came back" 2 of 4, "the thick nail" 2 of 5, the barrier 1 of 3, the clinic 1 of 2, the halfway softener 1 of 2 (NS-A06's first keep), the slow clock 0 of 3, runner's feet 0 of 3, the salon 0 of 2, the fine print 0 of 2. Across both rounds the photoreal-authority looks (news desk 3 of 3, rot organ 2 of 2, engraving 2 of 3, anatomy 2 of 4) outkeep the drawn ones. The launch: campaign120250102841410510"NailScience — NS-P2 Top 5 · wild statics R2 keeps · ABO $20/day", ONE ad set120250102843000510(his Oct 16 seven) at his $20, AD051–AD057, the Top 5 page at its April URL on the proven pair verbatim (same page and copy as round two: the statics are the only variable); 18–65 US, Advantage+ audience off, FB / IG / Messenger; created PAUSED 21:48 PT, ACTIVE at 00:00 Oct 30 PT on the same agent as round two, MGI and WAM. The account runs $170 a day from Oct 30, $20 over the plan's $150 cap on his call. Planbriefs/nailscience/launch/2026-10/NS-P2-STATICS-R2-ABO.json+.state.json. - 2026-10-30 (09:42 PT) — ROUND THREE PAUSED BY DAVID in Ads Manager (the account activity log: actor David Sanghera, Power Editor, campaign
120250102841410510Active → Inactive at 16:42 UTC) after day one's ~$22 on the seven r2 keeps, no orders. The agent had switched it on at 00:00:53 with round two (120250100703970510, ACTIVE at $50 CBO, ~$30 on day one, no orders yet). His call; not reversed — it goes back on only on his word, through adscaler. Twin v3.2 round 3 status + launch events, the doc's round-three update, the media plan board/page re-rendered. - 2026-10-30 (10:30 PT) — THE WINNERS, READ FROM META, AND ROUND THREE = VERSIONS OF THEM (David, in chat: "look at the history of the best winning ads on nailscience, list them here, and then lets generate version of those based on that").
tools/ad_codes.py report --brand nailscience --since 2026-04-01(110 ads with delivery): AD025 (W4) the bright-newsroom chyron 'PHARMACIST TESTED 7 TOENAIL KITS — ONLY 1 ACTUALLY WORKS' $1,318.82 · 33 buys · $39.96 · ROAS 1.08 (October alone $490.88 · 12 · $40.91) · AD024 (W3) the morning-sofa 'REVIEWED: TOP 5 TOENAIL KITS — THE WINNER MIGHT SURPRISE YOU' $438.10 · 17 · $25.77 · ROAS 1.65 · AD034 the newsroom EXCLUSIVE 'BIG PHARMA SUPPRESSED THIS NAIL FUNGUS FIX FOR YEARS' $114.75 · 4 · $28.69 · ROAS 1.39 · AD031 (W1) the standing segment 'PODIATRIST: WHY YOUR ANTIFUNGAL CREAM CAN NEVER WORK' $162.20 · 5 · $32.44 · AD033 the clinical tray of clipped fungal nails $399.18 · 7 · $57.03; then the agency cartoons AD014 (7 · $87.83) and AD013/W2 (2 · $113.91). The images matched by perceptual hash to the April library: AD025 =creative/nailscience/NEWS/NSR-NEWS-1.jpg, AD024 = NSR-NEWS-2, AD034 = NS-NEWS-9, AD031 = NS-NEWS-11, AD033 = NS-DIV-EVIDENCE-2 — four of the five best are the news-chrome frame: a serious female anchor, the gross toenail as a B-roll inset top-right, a kicker tab and the chyron headline, no product, bright set (his April notes: 'more of these types of styles', 'other women in this style, and also maybe today show', 'add the photo in the top right of the weird toe nail', 'brighter news room'; the kills: product shots, a doctor holding a bottle, a ranking chart). October r1/r2 agreed (the news desk kept 3 of 3). Round three = the exploit round: runbriefs/nailscience/statics/wild/2026-10-30-a.json(six situations on the winners' readers, 21 insets, the five winning sets + AD033's tray as formats; 28 slate rows = 7 core — the four winners' chyrons verbatim on new faces with FICTIONAL station bugs, plus three in their family — and 21 weird, the frame held and the line / inset moved: the fine print on the pen, the shoes, the bath mat, month two → month six, the sock snag, the hoof, sandal season, the halfway softener, the bead on wax, the $1,000 lacquer, the tray); every line chiefed (ledger batch 5),wild.py check0 fail; rendered 1:1 like the winners (aspectset on the shots) bytools/angle-statics-gen.py; the QA read at full size: one re-render before boarding (r3-10 drew a real network eye mark with real call letters, WJZ 13 — trademark = rejection → a plain KQXV 10 box). Board https://launch-center-hq.netlify.app/docs/nailscience/boards/statics-r3/ (anchor = AD025, need 7); askns-statics-r3-judge. On his keeps: round four on the media plan — NS-P2 on the proven pair, seven to an ad set, paused, on at midnight. Note the versions carry the April lines as the officer's inputs: r3-04 'Big Pharma suppressed' is not a Truth Gate claim (a verbatim port of his own AD034), r3-07 'NY podiatrist' is the page's persona (s5) — both flagged on the cards. - 2026-10-30 (15:42 PT) — ROUND FOUR LAUNCHED: THE FIVE VERSIONS OF THE WINNERS DAVID KEPT (David on the statics-r3 board 15:37–15:38 PT: keep r3-01, r3-02, r3-03, r3-27, r3-28, nothing killed, 23 left; at 15:38 he paused round two in Ads Manager — $50.33 on day one, no orders; then in chat: "okay lets relaunch in a campaign for $50"). Campaign
120250114416430510"NailScience — NS-P2 Top 5 · versions of the winners R3 keeps · ABO $50/day": ONE ad set120250114416880510at $50, AD058–AD062 (AD025's newsroom · AD024's morning sofa · AD031's standing segment, each chyron word for word on a new face · AD033's clipped-nail tray twice, now with a headline), the Top 5 page at its April URL on the proven pair verbatim, 1:1 images; created PAUSED 15:42 PT through adscaler (tools/ads_launch.py --go, planbriefs/nailscience/launch/2026-10/NS-P2-STATICS-R3-ABO.json+ state), read back (campaign PAUSED; ad set $50, Purchase, 18–65 US, Advantage+ audience off, FB / IG / Messenger; 5 ads in review) → agentcom.davidsanghera.ads-on-20261031-0000(+ sleeper) → ACTIVE 00:00 Oct 31 PT. The r3 read: the winners' own chyrons on their own sets kept 3 of 4 (the Big Pharma port left); the same frame with a new line 0 of 22; AD033's tray with a headline 2 of 2 → saved as frameworkclinical-evidence-tray(creative/dashboard/image-frameworks.json). The line that already bought is the asset: the next versions keep the winning lines and change the picture. The account, read from its activity log the same afternoon: round one's two campaigns have been off since 10-19 07:44 PT (Fotini Panagiotaropoulou paused both; §4) — the twin carried round one as LIVE until today; Fotini's 'Prospecting | US | Testing | CBO' ($150) and 'Prospecting | US | BFCM' ($100) run on the same Top 5 page since 10-29 → about $300 a day on the account from midnight. Twin v3.3 (round four inserted, the planned rounds → 5 / 6, round two paused, round one corrected, AD058–AD062, launch history + events, cap note) + doc (round four, calendar rows, decision 7) + board / page re-rendered; askns-statics-r3-judgeanswered; console update. - 2026-11-01 (21:20 PT) — "The new orders" read: nothing to apply. No NailScience order came in on Nov 1, and every order since Oct 26 registered in Meta on its day (David, in chat: "can you apply the new nailscience orders to the ads, it didnt register today for some reason"). The store (acm2): the last NailScience orders are #105842 Oct 31 14:16 PT $104.96 (tagged Testing CBO · 'IMG - Toenail fungus smiles') and #105777 Oct 30 15:49 PT $101.96 ('IMG - Before its too late'); none on Nov 1 as of 21:14 PT; the Shopify store has none since Oct 25. Meta's purchases equal the store's orders day by day Oct 26 → Nov 1 (5 · 1 · 0 · 1 · 1 · 1 · 0), the last two at the same values. The chain, checked the same night: webhooks 67 / 133 / 199 / 1 active with 0 warnings; /api/order-webhook answers an unsigned POST 401 and /api/capi an empty one 400; the Top 5 page → PDP → /checkout/?pkg=10 loads with the SDK cart at $59.98 on desktop and iPhone (tracking blocked for the check). Meta credits every NailScience purchase since Oct 22 to the campaigns Fotini runs (Sandbox CBO 10 · Ranking page 4 · Testing CBO 3 · BFCM 2 · Ranking page v2 2 · Sandbox CBO v2 1); rounds two, three and four have none. Why Nov 1 was quiet: at 01:18 PT Fotini cut 'Prospecting | US | Testing | CBO' $150 → $30/day and switched 'Prospecting | US | BFCM' off → $120.43 spent on Nov 1 against $232–$304 on the five days before; the pixel saw about six product-page visits and no add to cart (its stats count each event about three times). Round four after two days: $135.61 · 3 checkouts · 0 orders (AD058 $87.52, CTR 8.1%, 2 checkouts · AD061 $33.06, 1 · AD059 / AD060 / AD062 under $8 each). The midnight switch-on of Oct 31 fired (00:00:02 PT, campaign ACTIVE). David switched round four off at 21:11 PT and on again at 21:12 PT (Power Editor): it is on. Day-3 read Nov 2.
- 2026-11-02 — AN ORDER SUMMARY UNDER THE CHECKOUT BUTTON (ITEM + PRICE EACH): FOUR MOCK-UPS ON THE BOARD (David: "i think for all checkouts we need to add under the checkout button is the order summary, but it would just say the Item/ea price, show me a mock-up before introducing it"). Mock-ups only, nothing changed on any site. Four lanes (house rule: never one design), each rendered by injecting it under the real Complete Purchase button on the LIVE checkouts with the real cart line, price each = package price / units floored to the cent: A two-column line under Item / Price headings (minimalist-ui) · B soft card "Your order" (apple-design + emil-design-eng) · C one centred line "✓ item · $19.99 each" (impeccable) · D the package row echoed, "Your package" (taste-skill) — each on NailScience 3-kit, CellScience 4-bag, the NailScience trial ("Try Before You Buy — Monthly · $0 today") and NailScience desktop. Audit: all pass the 50+ floors (body 16–18px, labels 14–15px bold/caps, contrast 4.7–7.6:1); height on a phone A 94 · B 97 · C 44 (68 wrapped) · D 119px; no motion (checkout rule). Board
/docs/nailscience/checkout-summary/board.htmlon the Launch Center; askcs-order-summary; my recommendation A (the literal reading). After the pick: the winner goes intotools/checkout_selector.py(or the brand builders) for every brand checkout, staging → QA → production. The four lanes' CSS + markup live intools/qa/checkout_summary_mock.py(re-renders the captures:python3 tools/qa/checkout_summary_mock.py <out_dir>). - 2026-11-02 — ORDER LINE UNDER THE CHECKOUT BUTTON: DAVID PICKED B, THE SOFT CARD — LIVE ON EVERY CHECKOUT BUT SCULPTAPE THE SAME NIGHT (David, on the four lanes: "B"; my recommendation was A, the two-column line — he took the grouped card, the largest type of the four). A light-grey card right under Complete Purchase: "Your order", then the cart's item and its price each ("Moderate Fungus (3 kits) · $19.99/ea"; the trials "Try Before You Buy — Monthly · $0 today"). Built once —
tools/checkout_order_line.py(reads the 29next cart; units from the brand's rows, else the store's package quantity, else a "(3 bags)" count in the package name; add-ons like Shipping Protection left out; hidden until the cart holds a product; a value-guarded poll, no MutationObserver, no motion) andtools/checkout_selector.pylays it on every checkout it builds ("order_line": False opts out). Shipped to production 2026-11-02 night on NailScience, CellScience, SkinScience (all five), HumanFrame (checkout + echeckout), PeptideScience and Charm; Sculptape committed but NOT deployed (its deploy runs only from the /dev checkout — shared/cms). Every brand's diff = the card alone (Charm / Sculptape / PeptideScience got it straight from the module: their full rebuilds carry other undeployed changes). Proof:tools/qa/checkout_order_line_check.pyon staging or the built page over the live origin, then live — every lane PASS (NailScience 9/9 staging, 7/7 + 5/5 prod; tracking proof ALL PASS staging + prod; readability PASS 375/390; observer audit clean). Where a brand still has a full summary it now sits under the card: Charm + PeptideScience (the Total Order Summary bar right below), SkinScience + HumanFrame (the Item / Amount table further down) — flagged to David, nothing removed. - 2026-11-05 (09:50 PT) — ROUND FOUR READ: TWO ADS EARN IT ALL; THIRTY VERSIONS OF THEM ARE ON THE BOARD FOR ROUND FIVE, NOTHING LAUNCHED (David, in chat: "take a look at what's winning for the new nail science campaign you launched. Now let's do another round of $50 per day with some ad variations based off of the winning ones or the highest deep funnel metric ones. Send it over on the judgment and taste ads board so I can review them."). The read (campaign
120250114416430510, Oct 31 → Nov 5 morning, 7-day click + 1-day view): $325.34 · 5 orders · $65.07 an order · $329.89 on Meta, $368.86 in the store with add-ons; every purchase by click. AD058 (r3-01, the newsroom, AD025's chyron) $240.91 · CTR 6.18% · 156 page views · 30 PDP views · 16 add to cart · 16 checkouts · 2 pay presses · 4 orders · $15.06 a checkout · $60.23 an order. AD061 (r3-27, the tray) $59.75 · CTR 7.82% · 41 page views · 11 PDP views · 5 add to cart · 5 checkouts · 1 order · $11.95 a checkout · $59.75 an order: the cheapest step at every stage on a quarter of the money. AD060 $15.53, AD059 $5.11 and AD062 $4.04 barely delivered: no verdict on them. The store agrees order for order (#105908, #105941, #105974 Nov 2 and #106076 Nov 4 = AD058 · #106040 Nov 4 = AD061; #106076 carries the fbclid but landed with no ad tag, so its utm reads the prelander's fallback, 'sticky'). Who bought: all five 65+ (83% of the spend), all five on Facebook, four of five women. Honest bar: above the plan's $43 an order, and first-order revenue only just covers the ads; in the same window the rest of the account (Fotini's Testing CBO + BFCM until Nov 1) spent $341 for 1 order. The round (runstatics/wild/2026-11-05-a, board/docs/nailscience/boards/statics-r4/, 30 cards at 1:1): every card keeps one of the two lines word for word and changes only the picture, the reverse of round three, where a new line on a winner's frame kept 0 of 22 and a winner's own line 3 of 4. 14 on AD058's line, 16 on AD061's; 10 close versions (new anchors in their 60s for the 65+ buyer, seven crossed-out kits or the tray in the inset, half the screen or all of it the nail, a basin, a gloved hand, a black drape, AD061's line on the news desk) + the weird twenty (seven test tubes, seven dishes, the pharmacist at her counter, a phone photo of the TV, a reporter in the foot-care aisle, the nail on a video wall, the morning sofa, the toe wired up · a pill vial, the pill-counting tray, a petri dish under a loupe, an evidence bag, a paper towel by the sink, nine clippings in a row, the clippers mid-bite, a jar of it, crime-scene markers, a velvet ring box, a museum case, the tray on the news desk). He asked to judge on the board, so no slate stop (as r3). QA caught before boarding: the jar had 'Ball' embossed on the glass (a real trademark: re-rendered blank), the pharmacist's counter showed six kits for a line that says seven, the museum nail was too small to read, a white headline vanished on the black drape's light top. Next: his keeps (up to seven) →statics_board.py pull --round r4 --fill briefs/nailscience/launch/2026-11/NS-P2-STATICS-R4-ABO.draft.json→ one ABO ad set at $50, the Top 5 page, the proven pair word for word, created paused, on at the next midnight; he decides whether round four stays on beside it. - 2026-11-05 (10:20 PT) — THE ROUND IS JUDGED ON THE J&T BOARD, NOT THE LAUNCH CENTER PAGE (David, on the statics-r4 link: "This isn't the judgment and taste board that you put it in."). The J&T ads board is the dashboard: http://localhost:8769/dashboard/#nailscience-ads?round=4 (the board reads 'nailscience' in its brand dropdown, 'NailScience' is the April board; Rounds = Round 4; 30 cards with Keep / Refine / Kill).
wild.py boardhad already put the 30 there; my message pointed him at the Launch Center keep-or-kill page instead. His clicks save totools/jt/judgement-nailscience-ads-latest.jsonin the main checkout, and that file is the verdict, not the hub's log (statics_board.py pullreads only the hub). The Launch Center page stays up unlinked because it hosts the images the launch plan points at. The dashboard gained#<board>?round=N, which opens a board on one round. - 2026-11-05 (10:40 PT) — ROUND FIVE LAUNCHED: THE 27 VERSIONS DAVID KEPT, $50 A DAY, ON AT 00:00 NOV 6 (David, in chat: "Okay launch all the kept ones", after 55 seconds on the J&T board: 27 keep, r4-03 killed, r4-26 and r4-28 unmarked;
tools/jt/judgement-nailscience-ads-latest.json). Campaign120250214088960510"NailScience — NS-P2 Top 5 · versions of the R4 earners · CBO $50/day": four ad sets by what the card says, because 27 does not fit a seven-card ad set and he named $50 a day (the r1 and CellScience r7 precedent): newsroom line at the desk120250214090160510(7) · newsroom line out of the studio120250214420510510(6) · tray line, the clinical tray120250214496470510(7) · tray line, objects120250214598100510(7). AD063–AD089, the Top 5 page at its April URL on the proven pair word for word, 1:1, created PAUSED through adscaler; agentcom.davidsanghera.ads-on-20261106-0000+ sleeper flip it at midnight. Read back: campaign PAUSED, OUTCOME_SALES, daily $50; all four ad sets on Purchase on the pixel, US 18–65, FB/IG/Messenger, Advantage+ audience off, 7-day click / 1-day view, no budget of their own; all 27 ads with the exact headline, text and link (20 through review, 7 in review at 10:45 PT). The read: the winning lines on new pictures kept 27 of 30, against 0 of 22 for new lines on a winner's frame in round three: the words carry it, and the picture is free to move. The one kill is the only card that drew the line's claim as a graphic (seven crossed-out kits). Honest limits: about $1.85 a card a day, so Meta will back a few; his usual $50 an ad set would be $200 a day (asked). Round four stays on (no word to stop it), so the account runs $130 a day from midnight with Fotini's Testing CBO $30. r4-25 (clippers mid-bite) may not clear review. Meta's API quota throttled the launch (four 5-minute waits, 36 minutes in all); read-heavy jobs on this account wait for it. - 2026-11-05 (14:47 PT) — ROUND FIVE RAISED TO $100 A DAY BEFORE IT WENT ON; ROUND FOUR STAYS ON (David, in chat: "Do $100/day" and "Yes current one stays on"). Campaign
120250214088960510went $50 → $100 while still paused with $0 spent, so the +20% rule had nothing to protect. Read back from Meta: daily budget 10000 cents, paused, four ad sets with no budget of their own, all 27 ads waiting on the campaign, none back in review. About $3.70 a card a day. adscaler had no budget route and stays the only Meta mutator, so one went in first:POST /api/campaigns/budget {campaignId, dailyBudget}(whole dollars; any budget before a campaign's first dollar, then only raises of at most 20% a step; refuses a campaign with no campaign budget; writes nothing if the spend can't be read), deployed asadscaler-emz1sk86z. Before that deploy the live one's 79 source files matched the tree byte for byte, so nothing was dropped. The account runs $180 a day from midnight: round five $100, round four $50, Fotini's Testing CBO $30 ($150 of it this plan's, on its cap). Honest limits: the campaign's name on Meta still says "CBO $50/day" (adscaler has no rename route); only the name is out of date. Meta can spend up to 75% over on a day, $175 on round five.
- 2026-11-10 — VIDEO CONCEPTS, ROUND 1: TEN PITCHED, NOTHING GENERATED, DAVID PICKS (David, pasted into the session: "Develop 10 distinct video-ad concepts for https://trynailscience.com/products/nail-restoration-kit/ aimed at Women and Men suffering with NailFungus (45+). Use the attached product facts and references. For each concept, give me the hook, visual direction, and reason it fits this audience. Recommend one concept, but wait for my approval before generating. Do not invent product claims. Ask me if a necessary detail is missing."). Nothing was attached, so the Truth Gate (still open), §1/§3, the PDP's own lines, the twelve GO angles and the chief-PASS rows of the hook bank stood in. The pitch:
video-studio/scripts/nailscience/CONCEPTS-r1.md— ten concepts, each on a GO angle and a named bucket (seven of eight; trending left out, the news fetch is from Oct 29), every opener a chief-PASS bank row or a winner's line word for word: 1 the news desk, moving (NS-A03, AD025's chyron as AD058 ran it) · 2 the tray (NS-A03, AD061's line) · 3 it came back: the shoes (NS-A08) · 4 the click (NS-A09, the men's concept) · 5 the hoof (NS-A01, a real demo) · 6 after the shower (NS-A01) · 7 pool opens in March (NS-A10, city by state) · 8 ten toes (NS-A08) · 9 the pedicure excuse (NS-A05) · 10 the Army tin (NS-A11); plus the weird twenty as openers (not boarded). The fence held: no customer or clinician counts, no press, no before-and-after or time-lapse result, no doctor or pharmacist on camera, no generated demonstration, no named rival, no price or code in the video. Recommended: 1, because it is the only frame with repeat purchases here, round four's buyers (all 65+, all Facebook, four of five women) are who it speaks to, and it changes one variable (the picture moves); its line rests on the Top 5 page's pharmacist, open at the gate (s5) — if David holds that line, the pick is 4. The rap (V12) still waits for two more videos (his Oct 16 rule: three to an ad set); 1 and 4 would fill it. Honest limits recorded in the file: the news format is rare among video winners at large (about 14 of the 3,411 corpus winners), 2 needs real clippings or a new line, 5 needs a real shoot, 3 and 6 each need one fact on file before a script. Asks in §9.
- 2026-11-10 (evening) — THE VIDEO CONCEPTS RE-BASED ON RIVAL VIDEO RESEARCH; THE RECOMMENDATION MOVED (David, on the ten concepts: "Can we do it based on the angle studio because we did research for NailSCience on competitors right? Or no"). The honest answer: the angles yes (every concept sat on a GO angle and a chief-passed bank line), the competitor research no. What was on file was image ads (the Oct 18 sweep's 81 long-copy natives, still unjudged, and April's 30 statics); the bank held 0 rival videos and 0 rival transcripts. So the research was done the same day and put into the Angle Studio: a video-only Ad Library sweep ('toenail fungus', active, running 15 days or more: 150 ads, 35 advertisers, about $0.87 of the month's free Apify allowance), all 150 files pulled (135 distinct creatives), the first 90 seconds of the 90 longest runners transcribed, 86 frame sheets looked at, 10 off-market advertisers left out (116 creatives from 25 advertisers remain). The read (
ads/research/nail-fungus/RIVAL-VIDEOS-2026-11-10.md,video/notes.json): (1) the category's video argument is our own NS-A01 — creams sit on top, the fungus lives underneath — in the longest runs of Nuvira (188 and 177 days), maanufo (103 and 87), officialeureka (178); three rivals already say "shield"; (2) the closest rival is Nuvira, a pen with the same 25% undecylenic acid plus oregano, manuka and tea tree, six months on air; a creator video of theirs says "a two-step process" inside the one pen, so what is ours is the separate softening pen that opens the nail plate first; (3) the talking fungus is a cloned format (Nuvira 188 days, Dr. M. Bernstein nearly word for word, OnyGuard) and our rap V12 is the house version of it; (4) the most-produced format is the faceless macro of an ugly nail (Emily Parker: 30 creatives, to 205 days); (5) a home remedy is the standard cold open; (6) authority is a generated doctor, and none of the 86 creatives looked at uses a news desk; (7) creators tell one small story (Kerasal's "Pedicure Storytime" 102 days; nuubu's "If you stopped trying after 50, watch this" 155 days, five ids); (8) men enter by the work boot, and clippers "for seniors" sell the thick nail (99 days); (9) Approved Science (our acid) opened the boot-season angle 16 days ago. In the bank:seeds.mdcarries- refs:(49 rival videos tied to the twelve angles) and- signal:on every seed;sources146 creatives · 88 transcripts; verdicts, hooks and packets unchanged; video packets written for all twelve GO angles (angles/NS-A01…12/video.md). The slate, round 2 (video-studio/scripts/nailscience/CONCEPTS-r2.md; r1 superseded): 1 Everything you tried sat on top (NS-A01, new: the faceless explainer, the closest rival's 177-day format) · 2 the news desk, moving (less anchor, more nail) · 3 the tray · 4 the pedicure story · 5 the click (with the work boots) · 6 the shoes · 7 the honest clock (NS-A04, new) · 8 boot season · 9 ten toes · 10 the Army tin; the hoof and the shower tip left the ten. Recommended now: concept 1 first, the news desk second, both run with the rap — three formats in one ad set (the villain talks · the faceless explainer · the news desk) on the Top 5 page with the proven pair, so the test says whether the market's proof or our account's carries into video. Tooling:tools/angle-studio/rival_videos.py(new) +skills/angle-studio/SKILL.md §3c;angles.pynow reads a round-numbered manifest (a re-distil had dropped April's 30 statics from the evidence). Not done on purpose: no board rebuilt and nothing deployed (the angle board on Copy Sheet HQ still shows the Oct 30 evidence untilangles.py board+ the two deploys run, on his word); nothing scripted or generated.
- 2026-11-10 (night) — CONCEPT 1 PICKED: "go with concept 1 first" (David, in chat). SCRIPT 01 WRITTEN, GATED AND HEARD; ITS SHOT LIST WAITS FOR HIS KEEP / KILL / TWEAK; NOTHING GENERATED. Concept 1 = "Everything You Tried Sat On Top": angle NS-A01, a woman's voice over macro footage and drawings, no presenter. His go on the concept is not a render: the house rule for generated video is shot list first, judged before any still or clip is made. The script (
video-studio/scripts/nailscience/01_everything-you-tried-sat-on-top.md, round 2): written from the NS-A01 video packet and the live product page fetched that day, every line on the page's own words because the Truth Gate is open — the cast ("Rubbing cream on a toenail is rubbing antibiotics on a cast.", NS-A01-H9), "The fungus isn't on your nail. It lives underneath.", the roll call of what she tried (creams, tea tree oil, vinegar soaks), "You're not the problem. The products are.", "NailScience is two pens. The first pen is the step your creams skipped.", the second pen "goes through to the fungus … the strongest antifungal you can buy without a prescription", sixty seconds twice a day, the yellow fading in two to four weeks, the clear nail growing in from the base, the 30-day refund, then the tap. Five hooks, five buckets, all chief-PASS bank rows (H9 · H4 · H2 · H3 · H5). Gates: audit PASS (anailscienceblock added totools/audit-video-scripts.py: the 30-Day lock, CL-13 rival names and "Big Pharma" fail; the strength line, the result clock, the register word, the proof counts, the clinicians and the second person are reported for the officer); read-aloud in a mature woman's voice (GeminiGacrux): 131 words, 53.0 s spoken, 57 s with the pauses; the render transcribed back and every word landed. Both chiefs (COPYCHIEF-01.md): round 1 scored 7, eight changes applied as round 2 (the first pen named as the step her creams skipped; the second pen sent to the fungus; the refund before the tap; the page's "products are" line; three pronoun lines given their noun; "the sandals can come back out" cut as a spoken line and kept as a picture, because it promised sandals two breaths after "two to four weeks"); cap 8 while no customer, result photo or clinician may be shown. The shot list (SHOTLIST-01.md): 16 shots for the lead cut (12 generated stills, 8 short motion clips, 8 slow moves, 4 real product photos) + 4 opening shots for the other hooks; no face; one colour per pen (amber opens, teal goes through); the pens only from real photos or drawn the page's way; the one picture of a nail improving is a drawing. Honest notes for him: the pitch said 30 to 40 seconds and the script reads 57 (the closest rival's explainer runs 74); the pouch says "3 Pens" and the subhead says two, which decides shots 9 and 16; shot 11 stops when the teal reaches the fungus, where the page's own loop goes on to clear it. Tooling:tools/read-aloud-scripts.pyno longer stamps 2026-09-07 into the Duration row;skills/video-ad-scripts/SKILL.md§11 (a generated video with no presenter). - 2026-11-10 (late night) — THE SHOT LIST GOT A YES: "Yes, please use the best models to generate them" (David, in chat). VIDEO 01 IS MADE: FIVE CUTS, ONE PER HOOK, EACH IN 9:16 AND 4:5, ON THE J&T BOARD. NOTHING IS ON THE AD ACCOUNT. No shot was killed or tweaked. The models were picked from that week's public boards, then by eye on our own hardest shots: stills on GPT-Image-2.5 Sunburst (first on Arena's text-to-image and image-edit boards, 8 Nov; tested against GPT-Image-2.6 Sunburst and Gemini 3 Pro Image on the nail); motion on Seedance 2.5 for the hook's two halves and the two pens (first on Artificial Analysis, second on Arena; made beside Omni for those four shots and kept: the rubbing reads as rubbing and pen one's amber stays inside the nail) and on Gemini Omni 1.2 Flash for the other nine clips (first on Arena); the voice on Gemini 3.8 Flash TTS, voice Gacrux, one take (Eleven v4 ranks first and needs an ElevenLabs key this Mac does not have). Settled from the house's own records, not asked again: a kit is three pens, ONE softener and TWO antifungal (§1; the Top 5 page's own panels
products/nailscience/blog/ranking/assets/mechanism.webpandfeatures.webp); the video says "two pens" as the product page does and shows the page's own photo of one of each (hero-pack.jpg). Defaults taken where he gave no answer (each is a re-cut of the same clips unless noted): the length as written (the lead runs 52.7 s: the one-take voice reads faster than the 57 on the list); shot 11 stops when the teal reaches the fungus and the fungus stays green; no disclaimer on the video; no music. Where the cut differs from the list he judged: shot 14 is two shots (the yellow fades under "2 TO 4 WEEKS", the label leaves, then the clear band grows in as she says it, so the time label sits only on the page's own claim); shot 8 is a moving clip (the cabinet door closes); the oil drop is clear (amber is pen one's colour); the end card carries his subhead, a Tap below pill and the guarantee line, and no logo (the pouch carries the name). Hook 5 (the shower) is cut and HELD: "That's when medicine gets in" needs the kit's printed directions on when to apply. Checked before it reached him: every still and every clip looked at, each master sampled every second and its audio transcribed back word for word. Cost of the build: $15.35 across 92 paid calls (OpenAI $2.29 · Gemini $7.39 · KIE 1,134 credits = $5.67); the ledger is in the work folder. (The KIE balance read 2,636 credits after the build against 1,770 before it and 1,134 spent: 2,000 were added by someone else during the session.) Record:video-studio/scripts/nailscience/SHOTLIST-01.md(section Made),01.shots.json, the five01*.finish.json; enginetools/video-shots.py+tools/video-finish.py. - 2026-11-11 (07:37 PT) — THE THREE KEPT VIDEOS LAUNCHED, LIVE AT ONCE, STRAIGHT TO THE PDP AND STRAIGHT TO THE TOP 5 PAGE, $50 A DAY EACH (David, in chat: "lets launch the 3 new kept video ads from nailscience for $50/day straight to PDP and straight to Prelander, go ahead and use the best converting headlines and sharetext and targeting in the past and launch right now no need to wait"). His verdicts on the board (07:15–07:17 PT): video 01 hook 1 the cast KEEP, hook 3 the shelf KEEP, hook 2 the pen KILL, hooks 4 and 5 unrated; with V12 the rap (kept) that is the VIDEO group's three keeps. Shape: one ad set of the three videos per landing at $50 ABO (the 2026-10-16 rule: three videos to an ad set), so two campaigns and $100 a day; AD090 hook 1 · AD091 hook 3 · AD092 the rap, the same three codes on both pages. Copy = the proven pair, verbatim ("Pharmacist's #1 Nail Kit of 2026" + the "The yellowing, the thickness…" text): NailScience-only lifetime, the most Meta purchases (headline 74, text 79) and store orders (17 / 19), and the only copy that has sold on video (the animated video $24.29, the AI UGC video $44.48, both Top 5). Targeting = what every buying ad set ran: US, 18–65+, all genders, Advantage+ audience ON (April's winners W3 / W4 and the video that sold), Purchase on the pixel, no exclusions; placements manual Facebook / Instagram / Messenger (the 2026-09-21 HARD rule). Live at launch on his word ("launch right now no need to wait"), not the midnight switch. V12 runs trimmed: its second-to-last line, "Backed by 6,588 clinicians, 90-day guarantee!" (39.08–41.14 s, sung and captioned), is out — both pages and the 30-Day lock of 2026-10-14 say 30 days and the PDP says 658 clinicians; five beats at 129 BPM cut between the lines around it, re-transcribed and every frame read clean (
ns-video-cutclosed by the house trim rule; the original is untouched, the trim beside it increative/nailscience/videos/). Meta (read back 07:38 PT): PDP campaign120250299393720510/ ad set120250299393830510"NS-01 PDP · R1 · videos"; Top 5 campaign120250299394600510/ ad set120250299394650510"NS-P2 Top 5 · R6 · videos"; both ACTIVE at $50, Purchase, lowest cost, 7-day click / 1-day view, the targeting above; six ads ACTIVE and in review, each with its video, LEARN_MORE, its page and the pair verbatim. The account's only spend: David paused round five Nov 10 (5:20 PM PT) and the earlier rounds Nov 8, so these two are all of the $100 a day. Record:tools/jt/one-offs/nailscience-video01-2026-11/README.md. The PDP has had little direct paid traffic: four earlier ads sent there spent $330 for 2 Meta purchases, so the PDP leg is the test. - 2026-11-11 (09:58 PT) — A $50-a-day launch to a Carol Anne Foster post asked for; NOT launched: the post he named is gone (David, in chat: "Let's launch Nail Science for $50 a day to this post ID. 122129967033276653 - https://www.facebook.com/foster.carol.anne/posts/pfbid0DH3apoaLUpknAcdVFtN8E3KbbkzbgcvCD6SZxwSn7BTZk1Y2yEpJH4B8dLJi3Vvjl How does it work? Can you optimize for conversions for this one? Micro purchases?"). The post: logged out and logged in as David, Facebook says "This content isn't available right now"; the Page's own token reads today's ad posts but answers this ID with 'does not exist'; Meta's ad preview from the NailScience account = "Privacy Check Fail"; no ad in the 17 ad accounts the system user reaches ever ran it (its ID dates it to about Oct 29, 3 PM PT) — deleted. The one NailScience post on the page now:
927552797118674_122131182411276653, published 2026-11-11 09:40 PT, text only, "anyone know that toenail fungus thing celebrities are using to cure the yellow nails? i have a big fungus issue and need help", 0 reactions, 0 comments (no UPDATE comment, no link), and Facebook marks itis_eligible_for_promotion: false(the page's photo posts are eligible). The conversions answer: an ad that runs an existing post IS the post — the same words and picture, every like and comment lands on that one post, and the ad cannot add a button or a link to it. Purchase optimization needs the link on the ad itself, and so does every smaller website step (landing-page view, add to cart, checkout — the 'micro' events): a post whose link lives in the comments can be optimized only for engagement or reach (the 2026-09-20 'Call to Action Required' error;tools/jt-launch-post.pyruns these as POST_ENGAGEMENT / ON_POST, and its one round, Sculptape r8 2026-09-19, never spent). Buyers who click the comment link still count in the store, but Meta is not steering toward them. Three shapes put to him (§9). Truth Gate on the post's words: "celebrities are using" — no celebrity uses NailScience on file, so neither the answering comment nor a conversion ad can confirm it (CL-01). 10:15 PT: the post is back (David: "heres the post its avialable again", the same pfbid link). It is927552797118674_122129967033276653, posted 2026-10-29 10:07 PM PT, public: the same question as a text post on a purple background, 33 reactions, 32 comments; the author's UPDATE (Nov 9, 2:17 PM PT: "I ended up trying NailScience's two-step kit… Here's what I used: https://trynailscience.com/products/nail-restoration-kit" + a before/after photo of a toenail) and nine replies 4:01–5:02 PM that day lead the thread. Still not promotable: Facebook keepsis_eligible_for_promotion: falseand his own view shows 'See insights' and no 'Boost post' — a background text post cannot run as an ad of any kind (the page's photo posts can). The 09:40 copy of the question (122131182411276653, 0 comments) is still up. adscaler's launch route has no existing-post creative today either (every ad is built from a new creative). - 2026-11-11 (evening) — The first PDP-leg sale is not on Meta: the buyer's browser blocked Facebook's tracking; nothing on our side is broken (David: "nailscience has a sale today, can you see why it didnt post on fb"). Order 106403, 09:50 PT, $74.97 (Moderate Fungus, 3 kits $59.98 + VIP $14.99), desktop Windows Chrome, carrying AD090 on the PDP's tags (
utm_content120250299394340510, campaign 120250299393720510): a real ad click. Meta got the Purchase: the pixel's stats show one Purchase, SERVER, 17:00–18:00 UTC, match keys email · fn · ln · st · zip · country · external_id (the order webhook's relay). What it lacked is the click ID: the PDP URL arrived with no fbclid, and the order'sfb_fbcandfb_fbpare empty, so Meta's pixel cookie was never set in that browser — the mark of an ad blocker or privacy extension (it strips the fbclid and blocks fbevents.js). Without the click ID Meta can only match on name / email / address, which did not tie the buyer to the Facebook account that clicked; ~12 h later the campaign still shows 0 purchases (4 checkouts). Our side, checked: no trynailscience.com deploy since Nov 3 — Monday's order 106338 (Nov 9) on the same pages (via the Top 5 page, fbclid + fbc + fbp on the order) was credited to AD063 the same day; the PDP ad link re-run in headless Chromium with a test fbclid and every Meta / GA4 / CAPI beacon aborted (nothing reached any account): the address bar keeps the fbclid,_fbpand_fbcare set, Add to Cart →/checkout/?pkg=10carries the fbclid. Our count stands: the warehouse credits 106403 to AD090 by its tags (the PDP campaign's day one: $49.12, 1 order). Two gaps found on the way, nothing changed: (1) Meta gets only the kit's value — the Purchase goes onorder.created(09:50:08), before the VIP line joins the order (09:51:46), so Nov 9's $74.97 sale shows $59.98 on Meta; (2) no order carries a phone (the checkout asks for none — the open 10-26 call), andshared/functions/order-webhook.mjslooks forphone, not 29next'sphone_number, so a phone field alone would not reach Meta until the webhook's paths are fixed. - 2026-11-11 (night) — ANGLE STUDIO ROUND 2: TEN NEW ANGLES AND 70 HOOKS ON THE BOARD FOR HIS VERDICT; NOTHING ROUTED OR RENDERED (David: "Okay, now lets create some Angles, based on the Angle Studio / Hook-Buckets and all that, Give me 10 angles and concepts."). Round 1's twelve (all GO 2026-10-29) left five categories empty and the Nov 10 rival-video sweep showed lanes the market pays for that had no card, so round 2 fills both, one bucket each, all eight buckets walked: NS-A13 Saying yes when the grandkids want to paint your toenails (identity · micro-human · the granddaughter with the polish) · NS-A14 21 of 23 nail fungus patients improved on three softening ingredients (the one number, i4 Faergemann & Swanbeck 1989 · mechanism · the journal-page engraving) · NS-A15 Send the toenail kit back, even with the pens empty (offer, retargeting · competitor-ingredient · the 1970s print-ad seal 'EVEN EMPTY') · NS-A16 Sixty seconds twice a day instead of a 30-minute toenail soak (format · timing-usage · the split: the vinegar basin's 30:00 timer against the pens' 60 s) · NS-A17 Before the thick toenail has to come off (organ-outcome · long-term-fears · the removal tray) · NS-A18 The dark polish hiding your yellow toenail (hidden cause · environmental · the remover wipe that reveals the yellow nail) · NS-A19 The massage therapist who skipped the feet with the yellow toenail (story · micro-human · POV from the face cradle) · NS-A20 The thick toenail men blame on their work boots (the first men's card · environmental · inside the boot's toe box) · NS-A21 Heading to Florida in January with a yellow toenail (snowbirds · location · the news-desk 'SNOWBIRD ADVISORY') · NS-A22 The internet's toenail cures, tested on a cow hoof (failed solutions · competitor-ingredient · a real dye test on a hoof, filmed before any result is written). Evidence: the 92 Reddit quotes (problem words verbatim), the 2026-11-10 rival videos as refs (polish 74 / 71 days, nail removal 58, steel toes 32, the '15-second trick' cluster 15–22, home-remedy cold opens to 205), buyers.json (Florida, Texas, California among the top states by clicks) and a fresh trending pull (245 items, mostly SEO spam pages on university domains; the one usable item is this week's doctors' warning about nail salons, t52c818, on NS-A18). The Truth Gate is still open, so every claim id is a research claim; each card's research list sits in its notes. Copy-Chief (CL-21): headlines 7 PASS · 3 REWRITTEN (A14 named the condition, A19 named the toenail, A20 dropped a scene number) · hooks batch 6 + the weird twenty: 70 rows, 37 PASS · 33 REWRITTEN · 0 KILLED (18 fragments, 7 lines that dropped the toenail, 6 over 12 words); four graphic hooks are flagged for the officer, not softened (the removal tray, the toenail in a jar, the pliers, the tin snips). On the board: https://copy-sheet-hq.netlify.app/nailscience/?tab=angles (22 angles) and ?tab=hooks (250 hooks, batch 6 NEW); Copy Sheet HQ deploy
6af56070b7177ec03e50306f; askns-angle-board-r2. Found on the way, not changed (a separate task): the live refunds page carries CellScience's guarantee paragraph ('the fog… your kidney numbers') and the PDP has a 90-day guarantee block beside the 30-Day offer box; NS-A15 runs on the 30 days. - 2026-11-11 (night) — STEFANBRAIN'S VIDEO STUDIO IS THE ANIMATED LANE. IT HAS NO API, SO IT IS DRIVEN IN THE BUILT-IN BROWSER AND ITS RENDERS ARE FILED INTO OUR PIPELINE. NOTHING ON THE BOARD, NOTHING ON META (David, in chat: "Can you integrate StefanBrain as a way to generate videos as well?" · "Are you sure, StefanBrain has an MCP" · "Thats okay then, use it however you want. I suggest using Stefan Brain for more animated type stuff?" · "you can use your internal browser any time, i logged into stefanbrain, no worries right now keep doing what youre doing"). What was checked: the MCP is connected and working; its developer API and its MCP serve the same 41 tools and none makes a video (the live tool list, the public OpenAPI spec with 22 paths, and StefanBrain's own answer: "exclusively in the web application under the /videos studio"). The studio (
stefanbrain.com/videos): one prompt, 9:16, 4 to 240 seconds, 22 styles (Claymation, Pixar, Lego, 2D Cartoon, Comic Book, also Realism and iPhone), 12 formats (Songs, Music Video, Whiteboard, Problem → Solution, …), the account's product, voices on or off; billed in plan minutes (6 s = 0.10 min; 38.9 min left that night); output 720×1280 at 24 frames. It already holds 15 NailScience clips of his own (Video 1 to 15, made 2026-11-10 21:50 → 2026-11-11 21:44 PT, 4 or 6 seconds, Realism: the foot on a towel, the pens, the pouch, a pen on the nail), none on the board. The split (his suggestion, taken as the default): photoreal footage and diagrams stay ontools/video-shots.py's own engines; animated or stylized work goes to the studio. A real animated video still takes a script and a shot list and his yes before anything is generated. Built:video-shots.py renders(the studio's list, through the MCP) ·take(a downloaded render becomes a clip of a cut; the shot names"engine": "stefanbrain") ·drop(a whole video onto the J&T board). Proven on his Video 6: downloaded in the built-in browser (720×1280, 6.08 s, 690 KB), filed, picked up and decoded by the cut's own code; a file filed under the wrong label is refused. V12, the rap: its look (a 3D fungus monster in a cave under the nail, sung) and its file (720×1280, 24 frames) are what this studio makes, but the studio's list starts 2026-11-10, so where the rap was made is still not on record. Record:skills/video-ad-scripts/SKILL.md§11, "The animated lane". - 2026-11-11 (late night) — PRELANDER CONCEPTS, ROUND 1: TEN PITCHED, NOTHING DRAFTED; DAVID PICKS (David: "I wanted 10 concepts for prelanders, and the prelanders can include top 5 pages, 5 reasons pages, investigation, doctors, POV, quizzes, etc. 10 concepts based off our hook buckets and angle studio for prelanders"). The earlier reply had read "angles and concepts" as ad angles; round 2's ten angles stay on the board as candidates, and three of these concepts use them. The ten (format · angle · bucket): 1 Top 5 for thick toenails that catch on your socks (F4 · NS-A09 · environmental) · 2 5 Reasons your toenail fungus keeps coming back (listicle · NS-A08 · environmental; the statics' strongest situation, 5 of 5 kept) · 3 a foot doctor reads the back of every pen (F2 · NS-A02 · competitor-ingredient · Dr. Franks, PubMed screenshot in Chapter 2) · 4 the nail-softening step dermatology has used for decades (F3 · NS-A03 with NS-A14's 21 of 23 · mechanism · thesciencereviewjournal.com) · 5 how long will your toenail take to clear (quiz · NS-A04 · timing-usage; the result is the Mild / Moderate / Severe row,
?pkg=2|10|11) · 6 closed-toe heels at my daughter's beach wedding (POV story · NS-A05 · micro-human) · 7 my mother's toenails and the podiatrist who built the kit (F1 · NS-A12 · micro-human · Dr. Patel) · 8 snowbirds start their toenails in November (F5 · NS-A21 · location) · 9 five internet cures on a cow hoof, cut open (the test · NS-A22 · competitor-ingredient · a real shoot) · 10 I blamed my thick toenail on my work boots (POV, men · NS-A20 · environmental). Record:briefs/nailscience/prelanders/CONCEPTS-r1.md. His calls: which to build and the architecture of each (the four-row matrix, hook-enforced); one doctor per funnel (s5); the PDP's 'full clearance in a few months' against the gate's 9 to 18 months (p9); the hoof shoot; a pick on a round-2 angle is that angle's GO. Found on the way: the live 5 Reasons page (NS-P1) says '60-Day Money-Back Guarantee' (the law is 30; folded into the guarantee fix task with the refunds page and the PDP's 90-day block). - 2026-11-11 (late night) — PRELANDER CONCEPTS: ROUND 1 JUDGED IN CHAT, ROUND 2 PITCHED; THE TASTE RULE IS HIS OWN LINE (David: "Add
People put breast milk on toenail fungus. The fungus lives underneath.this is a good one, and #9. But alot of these are okay, but not very weird or catchy. See how the breast milk one is good. YOu need to use teh hook buckets to re-think the way these prelander angles will work"). Kept: #9, the cow hoof test, and his line as concept 1, verbatim. The rule his line sets: sentence one is a real, strange thing people actually do (a bucket, no product); sentence two is one plain fact that makes it pointless, and that fact is our argument (x1). A concept's headline IS the hook, never a description of the page. Round 2 (briefs/nailscience/prelanders/CONCEPTS-r2.md), one bucket each: 1 his breast-milk line (F5 · NS-A01 · trending) · 2 "Grandma's remedy for toenail fungus was peeing on it. She was half right." (F3 · NS-A03 · urine's urea is the softening ingredient) · 3 "People freeze their shoes to kill toenail fungus. Labs freeze fungus to keep it alive." (5 Reasons · NS-A08 · environmental) · 4 "The $1,000 toenail prescription comes in a bottle smaller than nail polish." (F4 · NS-A07) · 5 "I'm a foot doctor. My own toenail fungus beat every cream I prescribed." (F2 · NS-A01 · micro-human) · 6 "Halfway through my pedicure, she stopped and put on gloves." (POV · NS-A19) · 7 "My mother's big toenail fell off in the shower." (F1 · NS-A12 · long-term fears) · 8 "My toenail got so thick I cut it with tin snips." (POV, men · NS-A20) · 9 the cow hoof test, its five cures now picked by search data · 10 "How do you know if toenail fungus is dying? Look at the base of the nail." (quiz · NS-A04 · the question is +500% in search). Search data (Google Trends via StefanBrain, US, 5 years): Vicks is the most-searched remedy by far (average 59, peak 96); breast milk barely registers (0 on the same scale), so concept 1 says "people put", never "a trend"; rising: iodine +350%, castor oil +190%, "how to know if toenail fungus is dying" +500%, undecylenic acid +150%. Nothing drafted; his calls in the record. - 2026-11-11 (late night) — ROUND 3 OF VIDEO CONCEPTS: TEN ANIMATED AND TEN NOT. FOUR OF THE ANIMATED TEN ARE PICKED AND SCRIPTED, WITH A SHOT LIST EACH. NOTHING IS GENERATED: THE FOUR SHOT LISTS WAIT FOR HIS YES (David, in chat: "1. Yes" on the fungus talking from under the nail · "Lets come up with 10 video concepts animated and 10 non-animated concepts, just like we did before" · then, with three rows pasted from the angle board, "These 3 videos we can make, animated probably": A17 the tray, A19 the massage table, A20 the work boot). Angles: NS-A17, NS-A19 and NS-A20 are GO for video on that word. Logged in the bank only (
angles.py verdicts --from-file), not posted to the hub: he was judging on the board, and its verdict store overwrites on overlapping saves. Packets:angles/NS-A17/video.md,NS-A19,NS-A20. The other seven round-2 angles stay unjudged. Concepts (video-studio/scripts/nailscience/CONCEPTS-r3.md), animated: A1 the fungus talks back (NS-A01) · A2 the tray (A17) · A3 the massage table (A19) · A4 inside the work boot (A20) · A5 the freezer (A08: "I froze my shoes to kill the toenail fungus. It survived.") · A6 the Army training film (A11) · A7 the click (A09: "Grandpa's toenail clicked on the kitchen floor. Nobody said a word.") · A8 press your toenail (A03) · A9 the case of the fine print (A02) · A10 married 31 years (A05: "My husband has never seen my left big toe. Married 31 years."). Not animated: N1 the hoof (A01, a real test filmed as it happens; he kept the hoof test in the prelander round the same night) · N2 tin snips (A20) · N3 the vinegar bowl in the car (A16 ◇) · N4 one toenail painted black (A18 ◇) · N5 the dinosaur claw (A13 ◇) · N6 the excuse before the pedicure (A05) · N7 Mom's slippers (A12) · N8 nine months of pool season (A10) · N9 the news desk, moving (A03, carried from round 2) · N10 send them back bone dry (A15 ◇). The sixteen unpicked concepts were rebuilt the same night on the bar he set while judging the prelander concepts in another session ("alot of these are okay, but not very weird or catchy. See how the breast milk one is good"): every hook is a bank row shaped like his line, a strange thing people really do and then the plain fact that makes it pointless. ◇ = on a round-2 angle he has not judged: picking the concept is that angle's GO for video. Plus the round's weird twenty. Recommendation: the four picks as one family, the fungus first, then A5 the freezer; not animated, N1 the hoof if he or a creator can film it (one shoot serves the video and the prelander), else N2 tin snips. Scripts, round 2 (audit, read-aloud, the chief): 02 "The Fungus Talks Back" (49 s; round 3: its lead hook is his own line, word for word, "People put breast milk on toenail fungus. The fungus lives underneath.", said by the narrator and answered by the fungus; Vicks replaced the tea tree oil) · 03 "The Tray" (50 s) · 04 "She Skipped My Feet" (54 s) · 05 "Inside the Work Boot" (52 s, the first aimed at men).COPYCHIEF-02-05.md(7 to 8 at round 1, 8.5 to 9 after the fixes).SHOTLIST-02to05: nine or ten shots each and four other openings. Every product line is the live page's wording, read again that night. Looked up for 03 and 05, research items for the Truth Gate (his to rule): a thick fungal toenail left alone crumbles and commonly lifts off the nail bed, and the whole nail coming off is rare (two reviews on PubMed Central); taking a nail off without surgery softens it first with a 40% urea paste (DermNet, PubMed); shoes that make the feet sweat are a named risk factor, and men get toenail fungus up to three times as often as women (Mayo Clinic, a PubMed Central review; one regional study found no link to footwear). The studio test: Video 16 (6 s, Pixar style, voices off, 0.10 min): the fungus in a hammock under the nail while cream whitens the roof. 23 minutes to render. Filed as a take and sent to him; not on the board. Nothing on the board, nothing on Meta. - 2026-11-11 (late night) — HIS PICKS AND CALLS ON ROUND 2; FULL CLEARANCE = 3 MONTHS (David: "go with 1, 2, 7, 9, / #3 Yes / #4 generate the photos for them for now. photographer is working on getting them. / #5 3 Months is the full clearance, meaning your nail is fully healed. Use that and delete the rest of it. / Give me 5 more quiz concepts please"). Build: concept 1 (his breast-milk line, F5 · NS-A01), 2 (Grandma's remedy, half right · F3 · NS-A03), 7 (my mother's big toenail fell off in the shower · F1 · NS-A12) and 9 (the cow hoof test · NS-A22, GO by this pick, logged on the board as his). Doctors: two, as offered: Dr. Patel as the formulator in 7 (Dr. Franks for 5 if it is ever built). Photos: generated for now; the photographer's real ones replace them (concept 9's test photos included: the page does not take traffic on generated test results). The clock (Truth Gate, his word): b3 = full clearance in 3 months, the nail fully healed; p9 (the 9–18 month grow-out) deleted, out of every claims list and seed; the gate ask
ns-truth-gatere-posted with the same id. Quiz concepts r1 (briefs/nailscience/prelanders/QUIZ-CONCEPTS-r1.md): the nickel next to your big toenail · tap every cure you've tried · count the ways you hide one toenail · some toenails fall off in the shower · add up what you've spent on one toenail; every result picks the PDP's row and goes straight into checkout. - 2026-11-12 (00:30 PT) — THE FOUR PRELANDERS ARE BUILT AND LIVE (NO TRAFFIC YET) on thehealthreportdaily.com: NS-P4 /blog/breast-milk/ (his line, F5, NS-A01, FK 2.74) · NS-P5 /blog/grandmas-remedy/ (F3, NS-A03, FK 4.54; the 1989, 493-patient and 1998 studies attributed) · NS-P6 /blog/my-mothers-toenail/ (F1, NS-A12, FK 3.00; Dr. Patel's own PDP photo) · NS-P7 /blog/hoof-test/ (NS-A22, FK 2.83; review build: the results section says the shoot's photos go there, and nothing about what got through is written before the real test). Built by the new
prelanders/nailscience/build_article.pyon the NS-P3 chassis (head, tracking, footer disclosure and tail scripts byte for byte; gate before write: no dash, no 'mechanism', no deleted p9, no Trustpilot / real masthead / 95%, FK <= 7.06, no sentence over 24 words). Not carried from the chassis: the Trustpilot badge (s4), the 'As Featured In' mastheads (s3), the sidebar's '95% saw improvement'. Fixed on the way: the chassis's sticky button did not forward the click ids (it does on these four); an apostrophe in a tracking name broke the tail script on two pages (JSON-quoted now); the button-block line raised to 16px. Photos: 15 generated (gpt-image-2.5-sunburst, $0.61,prelanders/nailscience/_photos/r2.shots.json; no product rendered, no after-nail close-up, no hoof result) until the photographer's. Proof: desktop + iPhone render with every beacon blocked (5/4/5/5 images load, every button incl. the sticky forwards, 16px floor, no overflow, no script error); live, a test click on each page's first button reached the PDP with fbclid, utm_source, plid = the page, cta and eid_in. Deploytools/deploy-health-report.sh(now stages and smoke-tests the four; flow.json never ships), deploy 6af5721bd374d77986c252bd, every smoke path 200 with the pixel. Ledgerprelanders/nailscience/COPYCHIEF-r2-pages.md. Seen, not changed: the live Foot Doctor page (NS-P3) still says toenails take 9–18 months to grow out (his 3-month call), shows '95% saw improvement within 4 weeks · 89% … · 98% …' and a Trustpilot 3,214 badge, and its sticky button drops the click ids.
- 2026-11-12 (after midnight) — THE FOUR ANIMATED VIDEOS ARE MADE: 02 THE FUNGUS TALKS BACK · 03 THE TRAY · 04 SHE SKIPPED MY FEET · 05 INSIDE THE WORK BOOT. ON THE J&T BOARD IN 9:16, WITH A 4:5 OF EACH. NOTHING ON META (David, in chat, 2026-11-12 00:00 PT: "Yes, go ahead and generate all 4 videos. N5 should be animated. #1 yes #2 good #3 all good, keep going, forget the Truth Gate. #4. You can generate these." · later, when a download in the app's browser asked him to save: "oops, i cancelled it, just download it by default, dont ask me to save it"). What his answers settled: #1 a yes on all four shot lists, no shot killed or changed · #2 the read-alouds are good · #3 the three looked-up lines in 03 and 05 (a thick nail can crumble and lift; a clinic softens first; sweaty shoes) run as written · #4 the hoof (N1) may be generated · N5 (the dinosaur claw) moves to the animated lane. Two readings that are mine, not his words: "forget the Truth Gate" covers the claims in these four scripts as written and is not taken as closing the brand's Truth Gate (
ns-truth-gatestays open); and the hoof can be generated as an animated illustration, never shown as a test that was run. How they were made: each video's pictures are ONE whole film from StefanBrain's Videos studio (its planner keeps a character and a room the same across a film), filed as one clip and cut to the voice. 02: Video 17, Pixar style, 49 s, the studio's own voices for the narrator and the fungus, played whole. 03: Video 18, the Zack D Films format (a clean 3D look), 40 s, our voice (Gacrux), with video 01's two drawn clips for the yellow fading and the clear nail. 04: Video 19, Pixar, 50 s, our voice (Gacrux). 05: Video 20, Pixar, 49 s, our voice (Algenib). Each ends on the real product photo. What the films got wrong, and what was done: (a) 02's fungus said "antibacterial" for "antifungal" (five transcription models agreed): the line was re-voiced in the studio's editor (0.20 plan minutes) and joined into the film; all five now hear "antifungal". He also says "Your cream stayed" and "I'm packing" (the script: "creams", "So I'm packing"): left as said, the captions follow the voice. (b) 05's film ran both pens' liquid through the toe and out of the bottom: those frames are cut. (c) 04's film drew toenails on the soles and a yellow nail on both feet, where she says "one yellow toenail": the toes are framed out of that shot. (d) 04 shows her face once, where the list said never: kept, his to rule. (e) 03's pens carry "NailScience Co." and "Step 1" / "Step 2", legible and correct, where the brief asked for blank pens: kept, his to rule. (f) 03 was rendered twice: the first took 92 minutes; a second, in Pixar style, came back in 20 and is not used (it pushes past the tray and its liquid runs through the nail bed). Checked: every master second by second in both formats; every master's sound heard back against its script (every word; "NailScience" is heard as two words). Nobody has listened: the voices need his ears. Studio minutes: 4.11 used on this family across the two nights (the 6-second test 0.10 · four films 3.14 · the re-voice 0.20 · the second Tray 0.67); 34 left. Board:ns-v02-fungus(51 s) ·ns-v03-tray(50 s) ·ns-v04-massage(48 s) ·ns-v05-boot(52 s). Built along the way:tools/video-shots.pygainsscenes(where a film cuts),voice --from-file(a film's own voices as the take, heard back),splice(a re-voiced take into a film),"smooth"(a slowed span moves instead of stepping), labeluntilandoff_at, and a fix for a caption that came up 1.5 s early across a pause;tools/video-finish.pygains a shot's own 4:5 crop (anchors). Record:video-studio/scripts/nailscience/SHOTLIST-02to05(Made) ·skills/video-ad-scripts/SKILL.md§11. - 2026-11-12 (09:32 PT) — THE FOUR ANIMATED VIDEOS LAUNCHED PAUSED, ON AT 00:00 PT NOV 13: THE PDP AND THE TOP 5 PAGE, $50 A DAY EACH (David, in chat: "Launch Video 02,03,04,05 for NailScience $50/day using the prelander and PDP"). His launch word is the keep on all four (the J&T board cards stay unmarked), and they run as made: the voices, 02's three words, 03's branded pens and 04's one face shot (the four things §9 left to him). Shape = R1's (2026-11-11), the picture the only variable: one ad set per landing at $50 ABO, two campaigns, $100 a day; four videos to the ad set, one over the 2026-10-16 three-video rule, because his $50/day covers one ad set per page; AD093 video 02 · AD094 video 03 · AD095 video 04 · AD096 video 05, the same codes on both pages; the proven pair verbatim; R1's targeting (US 18–65+, all genders, Advantage+ audience on, Purchase, no exclusions, manual Facebook / Instagram / Messenger); the 9:16 masters. "The prelander" = the Top 5 page, as on R1 (his words then: "straight to Prelander"). Read in the account's activity log the same morning: David paused R1's Top 5 ad set "NS-P2 Top 5 · R6 · videos" on Nov 11 at 16:53 PT ($58.16, 1 checkout, no purchase); R1's PDP ad set stays on ($88.61, 8 checkouts on Meta, order 106403 ours). Told him in chat; a swap to one of the four new prelanders before midnight is his word away. Heard first: Whisper on the four final masters: 02 says "antifungal" (the re-voiced line holds), all four close on "a full refund for 30 days". Meta (read back 09:35 PT by id; the account's call limit, code 17, blocked the list reads after the launch): PDP campaign
120250315675400510/ ad set120250315675490510"NS-01 PDP · R2 · videos"; Top 5 campaign120250315677630510/ ad set120250315677830510"NS-P2 Top 5 · R7 · videos"; both campaigns PAUSED, the ad sets ACTIVE under them at $50, Purchase, lowest cost, 7-day click / 1-day view; eight ads ACTIVE and in review, each with its video, LEARN_MORE, its page and the pair verbatim. The switch: agentcom.davidsanghera.ads-on-20261113-0000plus a sleeper, scheduled from /dev. From midnight the account runs $150 a day (R1's PDP ad set $50 + these two). Reads: day 1 Nov 13, no verdicts; day 3 Nov 15 (python3 tools/ad_codes.py report --brand nailscience --pages --since 2026-11-13); keep at ≤ $43 an order or ≤ $4 a checkout, off at $60 with no order. Record:tools/jt/one-offs/nailscience-videos02-05-2026-11/README.md; plans + statebriefs/nailscience/launch/2026-11/NS-01-VIDEOS-R2.json,NS-P2-VIDEOS-R2.json. - 2026-11-12 (morning) — A RIVAL'S WHOLE AD PAGE READ FOR WHAT IT USES THE MOST: LUNAVIA (THE PAGE "SHANNON FITZGERALD"). THE PICK LIST AND THE BEAT SHEETS ARE WRITTEN; NOTHING SCRIPTED, NOTHING MADE; HIS PICK. David, with the page's Ad Library link sorted by total impressions: "Lets scrape this video ads library for themes and patterns of winning ads, and find the ads thats been used the most so that we can re-create it with my video editors." What was done: every running ad of the page swept on the free lane in Meta's own delivery order (Apify refused its token: 401, his to renew) · 111 ads → 171 files → 65 distinct videos → 23 ideas · every video transcribed and looked at · the brand's eight other pages found through its landing domain and read the same way (284 more ads, 102 videos): nine pages, 395 running ads, and no page shares a video with another. The rival: Lunavia, ONE pen with 25% undecylenic acid plus tea tree, oregano, manuka and propolis oils, "built by Dr. Kent Biehler, a podiatrist of 30 years", six pens for the price of three, 90 days; every nail ad lands on a 13-question quiz (fungusquiz.com, Heyflow + Checkout Champ). Its argument is NS-A01 (creams sit on top, a shield, open it first, then the acid), told as one pen that does both in order: the closest rival yet to the two-pen kit. The most used: (1) the toe models at the kitchen table, 3:25: places 1, 2, 6 and 12 in Meta's order, the only September video still running, copied into six more ads on Nov 7 and 8, and a second toe-model demonstration with the doctor in all ten video ads of his own page since Nov 11 · (2) four things you should never do, 1:45: place 5 in four days as one ad, and the Spanish page's biggest video · (3) three mistakes, 2:38: place 4, on since Oct 24 under four openers · (4) the bride's mother, 4:31, an acted film: place 3, on since Oct 24, four new openers in five ads on Nov 12 · then my receipts (place 7), the three villains, "Hey Jessica" (35 s). The patterns: one body under two to six openers (52 of the 65 videos; the new opener is 5 s on the median) · one caption and one headline under 109 of 111 ads · every body closes on the same nine beats · median length 109 s · the speakers are generated or acted and the small print says so · 72 of the 111 ads are two days old, and Black Friday is a frame over existing films (20 ads). The brand's oldest ads (409 and 296 days) sell the same pen for athlete's foot to young athletes on three other pages: not a nail comparable. My recommendation (mine, not his word): re-make the toe models first, because its key beat is "the right order" and two separate pens show that better than one; then four things. Not done, on his word: our own script on any of these (the angle is NS-A01; the hook round, the chief and the read-aloud come first), the Angle Studio's bank (these videos are not in
ads/research/nail-fungus/video/or on the seeds' refs), any board or deploy. Record:ads/research/nail-fungus/pages/shannon-fitzgerald/(RECREATE-2026-11-12.md,RIVAL-PAGE-2026-11-12.md,notes.json,page.json, the transcripts, the posters) andads/research/nail-fungus/pages/lunavia-network/. Tools:tools/angle-studio/rival_page.py(new),tools/fb-ad-library-page-sweep.py(--urls,--raw,listing_rank),tools/angle-studio/rival_videos.py(reads the free sweep too);skills/angle-studio/SKILL.md§3c. - 2026-11-12 (midday) — THE SEVEN MOST-USED VIDEOS OF LUNAVIA'S PAGE REMADE AS OUR OWN SCRIPTS 06 TO 12, EACH ON ITS REFERENCE'S CUT: SHOT FOR SHOT, BEAT FOR BEAT, AT ITS PACE, IN OUR WORDS. SEVEN SHOT LISTS WAIT FOR HIS VERDICTS; NOTHING FILMED, GENERATED OR ON META. David, on the morning's pick list: "Lets remake each (1-7) on close to verbatim in the pacing, scripting, shot list, and beats." How "close to verbatim" was read (my reading, not his words): the build is copied exactly and measured. Every shot of the reference has a shot of ours under it, the beats run in its order, and each beat of ours holds the reference's word count within 25%. The sentences are not copied: their script is theirs (CL-13), half of its claims are not true of our kit (one pen, ninety days, a named podiatrist, 500,000 users), and the house fences stand (no invented testimonial, no before-and-after, no invented authority A-03, every Truth Gate claim provisional). Where a reference beat hits a fence, the script carries the strongest line inside the ruling and the closer version is listed under "His calls" in that shot list. Wording closer than that is a per-line call on a script, on his word. The seven: 06 "Two Toes On The Table" ← the toe models (V01) · NS-A01 · 205 s · 70 shots · 624 words against their 622 | 07 "Four Habits" ← four things you should never do (V04) · NS-A01 · 105 s · 32 shots · 269 / 253 | 08 "Three Mistakes This Fall" ← three mistakes (V03) · NS-A10 · 159 s · 92 shots · 487 / 486 | 09 "Closed-Toe Heels On The Sand" ← the bride's mother (V02) · NS-A05 · 271 s · 116 shots · 678 / 660, an acted film | 10 "Pull The Receipts" ← my receipts (V05) · NS-A01 with NS-A02's small-print fact · 194 s · 75 shots · 605 / 571 | 11 "Three On Top" ← the three villains (V07) · NS-A01 · 66 s · 22 shots · 197 / 192, animated | 12 "Quick One, Ruth" ← "Hey Jessica" (V12) · NS-A01 · 35 s · 13 shots · 123 / 122. Every beat of every one is within 25%; the audit passes all twelve scripts. What stands where the references' fenced beats stood: a presenter with no profession where theirs is a foot-care worker of 28 years, a pedicurist or "an expert" · a clinic's order of work with no face where theirs is a named doctor (09 keeps a doctor as a character with no name: his call) · the Army in the 1940s (s8), a drawn calendar or the routine where theirs is a user count over a wall of customer clips · drawings with week chips where theirs are photos of toes getting better · the viewer's own receipts, in the second person, where theirs is one woman's story and her total · thirty days where theirs is ninety, with the reason it is enough: the yellow starts fading in two to four weeks, so she knows inside the thirty days (06, 08, 10) · no price said except "a month of it costs less than that bottle of oil" (10) · the small print on the back of a pen, "not effective on the nail" (c2), where theirs is a doctor found on a phone. Three lines were looked up and go to the Truth Gate as research items: a salon may refuse an infected nail and some state boards require it (South Carolina's rule, California's board, Texas's consumer guidance; the scripts say "some states") · nail fungus spreads on shared towels, mats and shower floors (the dermatology academy's patient pages) · a fungus feeds on the nail it lives in. The chief (
COPYCHIEF-06-12.md), applied: five of the seven never said "NailScience" out loud, 09 for four and a half minutes; it is now said in all seven where the relief arrives · 06 gives the order of work before it names the pens, as its reference does · 10 no longer tells her she read every label twice and then asks her to find the small print · 10's joke about oils that smell like a spice rack is gone (oregano and clove oil are in our second pen) · 09 ends on "In the car. Both pairs." · 12's five test questions all fit Ruth's one answer. Scores after the fixes: 8.5 to 9. The read-alouds (email/nailscience/video-scripts-audio/06.mp3to12.mp3) were transcribed back: 93% to 99% of each script's words are heard, and the misses are numbers written as digits and the word undecylenic. The render's voice is slower than the references' speakers, so six of the seven run longer than their cut (12 the most: 47 s for a 35 s cut). Nobody has listened. Built:tools/video-remake.py(new:fit,shotlist,check) ·tools/angle-studio/rival_page.py shots(a reference's shot log and sheets) · the audit'sPacerow and its note when a script never says the brand · the read-aloud'sVoicesrow ·skills/video-ad-scripts/SKILL.md§12. Corrected: the morning's pick list said the bride film never states an offer; its last shot is the product under a 90-day guarantee label. Not done, on his word: anything filmed or generated, a brief to the editors, the seven sets of five hooks onto the Copy Sheet (CL-23), any board or deploy, anything on Meta. Record:video-studio/scripts/nailscience/(scripts 06 to 12,NN.remake.json,SHOTLIST-06to12,COPYCHIEF-06-12.md, README) ·ads/research/nail-fungus/pages/shannon-fitzgerald/shots/. - 2026-11-12 (13:55 PT) — THREE OF LUNAVIA'S VIDEOS, CUT WHERE ITS PEN COMES IN AND CLOSED ON OUR OWN PENS, LAUNCHED PAUSED TO THE PDP AT $50 A DAY, ON AT 00:00 PT NOV 13 (AD097–AD099). David, in chat, in order: "Launch these into another $50 campaign for tomorrow" (five Lunavia Ad Library downloads) · "Lets launch them for $50/day spend so we can see which one works and which one to remake" · "Lunavia is MY BRAND also" · "do the cut versions for tonight". Why cut, not as made: each of the five sells Lunavia's offer by name (the Lunavia pen, Dr. Kent Beeler, 500,000 Americans, buy three get three free, up to 50% off, code PROTOCOL, a 90-day guarantee); a NailScience click gets a 30-day guarantee and none of it. The cuts: AD097 the toe models (
1731414118155274, Lunavia's most-delivered video) kept to 78.26 s, "…the right killer's been chained up outside the house it's meant to guard."; AD098 three mistakes after summer (1456664803030634) and AD099 the pedicurist in fall (1148128360984317, the same body) kept to 70.10 / 70.40 s, "…the biofilm slime that the fungus lives inside of."; then video 01's own close (21.03 s to the end, "NailScience is two pens…" to the product card, the 30-day refund, "Tap below"). Heard back with Whisper and read every second with OCR: no Lunavia word, doctor, number, offer or formula left. Not cut: the polish reply (the offer from its first line) and the drawn story (its hook promises bees, its second half is propolis, not in our pens). Shape: one campaign, one $50 ABO ad set of three (the 10-16 rule), the PDP (he named none; the PDP carried R1's order), the proven pair verbatim, R2's targeting; campaign120250317776060510, ad set120250317777940510, read back by id; the midnight switch joined the R2 agent. Flagged to him, still in what was kept: the toe models' "from this to this before the holidays" + the label "Watch your toes clear out by the holidays" (his 3-month clearance; script 06 says spring) and its digital spokesperson's "28 years I did feet for a living"; the pedicurist voice under "Actor portrayal". From midnight the account runs $200 a day. Record:tools/jt/one-offs/nailscience-cuts-2026-11/README.md. - 2026-11-12 (afternoon) — THE SEVEN REMAKES ARE BEING MADE: 11 IN THE ANIMATED LANE, THE OTHER SIX BY US WITH A GENERATED PRESENTER, AND A QA AGENT CHECKS EACH AGAINST ITS REFERENCE AND ITS SHOT LIST. 11 AND 12 ARE BUILT; 06, 07, 08, 09 AND 10 ARE BEING BUILT. NOTHING ON META. David, in chat: "generate video 11 in the animated lane, rest lets render them." Then, while it was under way: "create a QA agent that will compare the two videos to make sure they are similar beat by beat and look at the shot list to make sure its what was intended for". How it was read (mine): his word answers the shot lists' "who makes each" (no editors, no actors: the studio for 11, our generation for the rest). The other calls in the shot lists (the doctor in 09, who speaks, a price, the customer count) are built as written and stay his. The engine, picked by a test on the hardest shot: the same ten-second line of 06's presenter on Gemini Omni 1.2 Flash (33 s to make, $0.70, every word, 2.98 words a second, the reference's pace) and in the studio's iPhone style (about half an hour). Omni won on control: exact words, a locked first frame, the length asked for. It takes no sound in, so each presenter's voice is the clip's own. 11 "Three On Top": one Pixar film from the studio (Video 22: 85 s asked for, six character voices, every line as written, 1.42 plan minutes), its quiet between lines shortened to the reference's nine beats at natural speed: 66.2 s against the reference's 65.8, every beat within 10%, all 197 words in place. The signs came back blank as asked and the words were laid on them. 12 "Quick One, Ruth": the presenter of 06 at a desk in four talking clips, the split screens made of video 01's drawings and clips and real product photos: 35.3 s against 35.0, every beat within 3%, $6.82. Masters (9:16 and 4:5):
creative/nailscience/videos/11-three-on-top/finish/and…/12-quick-one-ruth/finish/. Found on the way: a talking clip given too many words drops a clause, and it drops the hardest one ("twenty-five percent undecylenic acid", twice) · one take in about ten comes back in another accent (British, Australian): a listening model now judges every clip three times and the lowest score counts, because it is not steady itself · the studio's voices leave long quiet (49 s of speech in the 85 s film). Built:tools/video-shots.pytalk · voices · presenter · matte (the Mac's own Vision framework) · tighten · split-screen, tile and cut-out layouts · a bank of kept pictures shared between videos;tools/video-remake.py qa; the QA agent.claude/agents/video-remake-qa.mdwith/video-remake-qa;skills/video-ad-scripts/SKILL.md§13. Nobody has listened: every voice is checked by transcript and by a listening model only. Record:video-studio/scripts/nailscience/NN.shots.json,NN.finish.json,QA-NN.md.
- 2026-11-12 (night) — 11 OPENS ON THE PROBLEM NOW: HIS NOTE ON THE FIRST CUT, AND THE SECOND CUTS OF 11 AND 12 SENT TO HIM. NOTHING ON META. He watched 11 and 12 and wrote, on 11: "The problem with the animated one is, it doesnt show the initial problem, like Youve got fungus in your nail, its festering under the nail, and then go into the rest of the video, you gotta setup the problem statement at the beginning". Done (11): a 7.6-second opening the reference does not have. The narrator (Gacrux, video 01's woman) says "You've got a fungus under that toenail. It's been festering there for years. And it's got three best friends in your bathroom." over the fungus asleep in his cave under the nail while green threads creep across the floor and into the nail (Gemini Omni, made from the film's own frame, so he is the same fungus), then the bathroom; the cream now speaks 0.4 s into his shot. Chiefed (
COPYCHIEF-06-12.md, round 2, 8.5). The QA's fixes went in with it: the fungus stays in his hammock while pen one opens the nail and pen two's teal reaches him (both shots made again on Seedance from the film's frame: the film had lost him), the wrongly drawn foot behind the polish framed out, the labels out of the top 14%, the sign words centred, the fungus's captions on his voice. 73.4 s against the reference's 65.8 (7.6 s is the opening); all 219 words in place; every beat within 10%. Done (12), on QA-12: the 4:5 laid out on its own (the splits had lost the pens), Ruth's clip joins frame-exact and on cuts (a punch-in where one fell mid-shot), PEN ONE / PEN TWO tagged in amber and teal, plain white captions, the small print the script asks for ("AI-generated portrayal."). 35.3 s against 35.0, every beat within 3%. Omni's daily quota (200 requests) ran out at about 21:47 PT and frees about 00:05 PT: 06 to 10's builders finish what needs no new talking clip; 09 waits for 33 of its 60 clips. Still his: his ear on the voices (nobody has heard them; "I'm your polish" at 0:08 of the first cut was misheard three ways), the fungus's design (11's own, not 02's), and the calls built as written. - 2026-11-13 (small hours) — ALL SEVEN REMAKES ARE MADE AND SENT TO HIM. THE QA AGENT CHECKED SIX: ALL PASS WITH FIXES, AND EVERY FIX LIST WENT INTO A NEW CUT. 09 IS IN QA. NOTHING ON META. Made: 06 (3:25.1 against the reference's 3:25.1, 624 of 624 words, $20.57), 07 (1:45.4 / 1:45.3, 269 / 269, $10.60), 08 (2:40.7 / 2:38.6, 487 / 487, $17.20), 09 (the acted film: 60 talking clips on six people, 116 shots, 4:30.5 / 4:31.2, 676 of 678 words heard back, $38.17, "AI-generated dramatization." at the foot, the doctor in as written), 10 (3:14.4 / 3:13.6, 605 / 605, $25.94); 11 on its second cut, 12 on its fourth. $122.31 of generation for the seven. What the QA agent found, and what was done: a talking track that ran short of its timeline, so her picture jumped a few frames before cuts and her voice drifted ahead of the captions by up to 0.69 s (06, 07, 08, 10, 12): the presenter join now fits picture and sound as separate parts, each exactly its slot, the sound checked to the sample · split screens the shot lists had coded full screen (06: ten shots; she was off screen 79 s against the reference's 45 s, now 45.5 s) · two right feet with a toe ring (07) · words Omni drew into a clip (08 t10, made again) · product photos that show two Step 1 pens (
hero-angle,m-flat-lay,l-nightstand): every Step 1 and Step 2 moment now useshero-packor the bundle photos · clips whose voice measured off pitch (06 t18, 10 t11 and t29, made again;voicesnow measures every clip's pitch) · 4:5 masters that cropped the pens out of the splits or squashed full-frame shots (06 and 12 compose their 4:5 on their own; the squash was a tool bug, fixed). Still his: his ear on every voice (nobody has heard them), the doctor in 09, the generated hands holding plain pens (10 shot 56; 09 shots 80, 82, 83), the music bed at upload, and the shots that still differ from their lists (eachQA-NN.md). Record:video-studio/scripts/nailscience/README.md(the made table),QA-06.mdtoQA-12.md,NN.shots.json. - 2026-11-13 (03:20 PT) — 09'S SECOND CUT IS MADE AND SENT TO HIM, ON QA-09 (PASS WITH FIXES: OK 93, CLOSE 11, MEANT 3, OFF 0, WRONG 9). ALL SEVEN ARE NOW ON A CUT THE QA AGENT'S FIXES WENT INTO. NOTHING ON META. Done: shot 1's 4:5 shows the bride's face (anchor 0.0 for 2.8 s) · the friend's lips move through her lines in 72 and 105 (t0807 and t1006 made again; heard word for word, same speaker, pitch +3% and +7.6%) · 79 is the pack shot with a Step 1 and a Step 2 pen and 91 is
bundle-1(both were two Step 1 pens) · 68's malformed hand made again · 51's captions off her chin · the beach photo in the frame he hangs in 113 and 115 · 4:5 anchors for 6 and 93 · 55 relit as a grey morning · the wedding sound now plays at its own speed and stops on the cut into the hallway, so the last 11.8 s are silent. 4:31.4 against 4:31.2; 678 of 678 words heard back; numbers pass; $39.14 for 09, $123.28 for the seven. Not done: the pens in the friend's hand in 79 (no allowed photo shows a hand holding the pair, and the product is never generated); the day-chip progress bar. Still his: the caption style (the reference's is yellow italic), a bed under the film (it falls near silent between lines about thirty times), the doctor, the voices. - 2026-11-14 (afternoon) — NO AI LINE ON ANY VIDEO. "AI-GENERATED PORTRAYAL." CAME OFF 06, 07, 08, 10 AND 12, AND "AI-GENERATED DRAMATIZATION." OFF 09, IN BOTH SHAPES. NOTHING ON META. David, in chat: "Remove all the "Ai-generated portrayal." at the bottom of all of the videos, our legal council is going to be reviewing and making sure all of the compliance is up to date. We dont want you to do it." The rule: we put no small print on a video. Legal counsel reviews every video and sets any. A script's Disclaimer row reads "none from us", and no config carries
finish.disclaimer. We had added the line on our own judgment; 01 to 05 and 11 never had it. The masters were rendered again from the same finish specs with only the line taken out, and the review copies on Video Review HQ were replaced in place, labels kept. - 2026-11-16 — HIS FOUR NOTES ON VIDEO REVIEW HQ ARE IN THIRD CUTS OF 07 AND 09. 07 SHOWS THE FUNGUS SPREADING TO THE OTHER NAILS AS GREEN POWDER; 09 OPENS ON THE BEACH WITH HER SICK TOE IN PLAIN VIEW. NOTHING ON META. His notes, on the second cuts: 07 shot 13, "This partof the video should show the fungus spreading to the other nails." and "Show more fungus-type style like spores on the different places the fungus lives. Green powder spores." (and Needs changes on the cut); 09 shot 1, "Why is it in a bridal shop and not at a beach? It doesnt make sense?"; 09 shot 4, "It should show the toe that has fungus on it." Done: 07 shot 13 is the same burgundy-polished foot; green powder creeps from under the polish and spreads onto the second and third toenails, which turn yellow (a 4 s Omni clip between two new stills; Gemini's filter blocked the words "fungal spores", so the motion is described as green powder), then holds on the spread; in 4:5 the captions over it sit lower so the streak shows. 09's whole opening scene (shots 1–6) moved to the beach where the wedding will be: six new stills, the five lines taken again word for word, the same two women; shot 4 is her bare foot out of the navy heel on the sand, the thick yellow big toenail in plain view, mirrored so the nail is on her right foot as in every later shot; shot 1's 4:5 crop keeps both faces and her heels in the sand. QA-07 and QA-09 third cut: PASS WITH FIXES, every fix done; the numbers pass and nothing outside the changed shots moved. Cost: 07 +$0.53 ($11.13), 09 +$1.89 ($41.03). Still his: the voices (Becca's first line reads about 5% higher than her car lines; at 1:14.6 an extra "on" was heard, unchanged since the first cut); a music bed for 09's silent last 11.8 s; the gold caption accent against pen one's amber (07, open since its first QA).
7 · Asset map
- PDP source of truth:
products/nailscience/index.html(177 KB · 53 images · 15 videos; hero set,ns-mechanism.gif, review photos inproducts/nailscience/assets/, originals inassets/_originals/). Collection pageproducts/nailscience/collection/(built 2026-06-01, never deployed). - Creative:
creative/nailscience/— 81 J&T images (DOC 30 · NEWS 22 · DIV 13 · TIMELINE 6 · CON 3 · QUAD 2 · COMPARE / CHART / CONTRAST / CONTR / EXPERT 1 each) +manifest.json+pages-manifest.json+ the5reasons-v3/master.jsonboard. No hero / gallery / bundle-card round was ever shot for the buying set. - Ad copy bank:
skills/dtc-ad-templates/source-ads-nailscience-2026-04.md(headlines + ADCOPY 1…; port the proven lines verbatim). - Research:
tools/deep-researcher/runs/nailscience/at-ns/07_copy-brief/response.md— the April copy brief, the research base (Mode A; David 2026-10-14: no deep-research re-run). Notruth-gate.md, noangles/, no Visual DNA yet. - Logistics:
logistics/nailscience/. Profit:profit/nailscience/(May 7–31). Economics:tools/proforma/assumptions/nailscience.json→tools/unit-economics/data/nailscience.json. - Checkout template variant:
template/checkout/02/nailscience/(css + images incl. the 30d/90d badges and reviewer photos). - Email:
email/nailscience/subject-lines/manifest.json(empty shell); the real bodies are in Klaviyo. - Vault:
~/Downloads/davidsanghera/Projects/NailScience/. - Video 01 "Everything You Tried Sat On Top" (made 2026-11-10): the record is
video-studio/scripts/nailscience/(01.shots.json, five01*.finish.json,SHOTLIST-01.md); the media iscreative/nailscience/videos/01-sat-on-top/(26 stills, 13 clips kept, the voice takes, ten masters infinish/; not in git, a copy in the main checkout); the five 9:16 masters sit increative/nailscience/videos/asns-v01-h1-cast…ns-v01-h5-showerwith sidecars, which is what puts them on the J&T board (localhost:8769/dashboard/#nailscience-ads, Group VIDEO, beside the rap). The video kit (the product page's DM Sans as TTF, the pouch's navy and gold, amber for pen one, teal for pen two):video-studio/fonts/nailscience/and thenailscienceblock intools/video-finish.py. - Videos 02 to 05, the animated family (scripted 2026-11-11, made 2026-11-12): the record is
video-studio/scripts/nailscience/(scripts02_the-fungus-talks-back.md,03_the-tray.md,04_she-skipped-my-feet.md,05_inside-the-work-boot.md; aSHOTLIST-NN.mdeach with Made at its foot;NN.shots.jsonandNN.finish.jsoneach;COPYCHIEF-02-05.md; the concept listCONCEPTS-r3.md). The media iscreative/nailscience/videos/02-the-fungus-talks-back/,03-the-tray/,04-she-skipped-my-feet/,05-inside-the-work-boot/(the studio film and its record inclips/, the downloads, the voice take, the raw cut, two masters infinish/; not in git, a copy in the main checkout). The four 9:16 masters sit increative/nailscience/videos/asns-v02-fungus.mp4,ns-v03-tray.mp4,ns-v04-massage.mp4,ns-v05-boot.mp4with a sidecar each: the J&T board lists that folder. The studio's films are its Video 17 to 21 (21 is the unused second Tray). - Rival page teardown, Lunavia (2026-11-12):
ads/research/nail-fungus/pages/shannon-fitzgerald/—RECREATE-2026-11-12.md(the pick list and the beat sheets, for David and the editors) ·RIVAL-PAGE-2026-11-12.md(the read, the launch calendar, all 23 ideas and 65 videos with their Ad Library links) ·scripts/(what every video says, timed) ·video/(local, not in git: all 171 files, the seven 720p referencesV01-the-toe-models.mp4toV12-hey-jessica.mp4, the frame sheets invideo/sheets/). The brand's eight other pages:ads/research/nail-fungus/pages/lunavia-network/. - The seven remakes, scripts 06 to 12 (written 2026-11-12):
video-studio/scripts/nailscience/— the scripts06_two-toes-on-the-table.mdto12_quick-one-ruth.md, a mapNN.remake.jsonand a shot listSHOTLIST-NN.mdfor each,COPYCHIEF-06-12.md. The cuts they are fitted to:ads/research/nail-fungus/pages/shannon-fitzgerald/shots/(V01, V02, V03, V04, V05, V07, V12: one row per shot with its times and the words said over it), with a sheet of every shot's first frame invideo/sheets/<cut>-shots-pNN.jpg(local, not in git). Review audio:email/nailscience/video-scripts-audio/06.mp3to12.mp3. Tool:tools/video-remake.py. Nothing is made from them yet. - StefanBrain's video studio (the animated lane, 2026-11-11): what it has made is listed with
python3 tools/video-shots.py renders(15 NailScience clips of David's that night, Video 1 to 15, and the wiring test, Video 16). Its files live in the app until downloaded. The wiring test's files:creative/nailscience/videos/sb-studio-test/(media, not in git, not on the board). - Media plan 2026-10:
briefs/nailscience/media-plan-2026-10.json(the twin: pages NS-P3/NS-P2/NS-P1, ads W1/W2/W3/V12 with the pair that bought, bench, rounds, rules, the turn-on sequence, the destination fix, preflight) +.md(the plain-language plan) → renderedtools/brand-hub/docs/nailscience-media-plan-2026-10.html; posterstools/brand-hub/docs/nailscience/media-plan/{ads,candidates,pages}/; V12 hosted at the UGC Studio (docs/nailscience/media-plan/video-index.json). Launch boardtools/launch-board/nailscience.json. Ad codesbriefs/nailscience/ad-codes.json. Deploy:tools/deploy-nailscience.sh(gated, with functions, smoke;--draftfor a preview). The video:creative/nailscience/videos/video12_edited.mp4(+.jsonsidecar; moved out ofproducts/nailscience/ads/so it never deploys).
8 · Live surfaces (checked 2026-10-13 · staging added 2026-10-14 · PRODUCTION = the relaunch build since 2026-10-15 11:20 PT)
STAGING (David judges here; noindex header): https://nailscience-staging.netlify.app — Netlify site nailscience-staging · id 9389f064-be0e-4e46-aa38-2a37fb53b9e4 (EcomScalers, created 2026-10-14). Deploys ONLY through products/nailscience/deploy.sh (staging default · --prod = trynailscience.com; hygiene gate + --functions + smoke). Phase 4 build on staging 2026-10-14: the PDP at /products/nail-restoration-kit/ (moved from /; the home takes /; old /?pkg= and /?ob= links 301 to the PDP) with the house radio offer box on Ladder A + the 14-Day Free Trial row (29next package pending), the 30-Day Money-Back Guarantee card (pdp_house), the site footer; the homepage (lane C "Product first", ported from CellScience); /collection/; the seven legal pages at /pages/<slug>/ (one-sided stack rewritten for a topical kit: Product Disclaimer + Shipping overridden, refunds = "send the kit back, even empty"); robots.txt + the sitemap index derived from the pages manifest. Gates: check_compliance.py PASS · readability PASS (PDP FK 5.4, home 2.7) · deploy hygiene PASS. Builders: site_chrome.py · pages_content.py · build_site_pages.py · check_compliance.py · wire_pdp_footer.py · tracking_head.py (GTM) · pagespeed.py + fonts/ · build_pdp_read50.py (the 50+ readability pass — run LAST after any PDP rebuild); copy ledger COPYCHIEF-pdp-r1.md. Rebuild order for the PDP: offer_box_v2.py nailscience <pdp> → pdp_house.py → wire_pdp_footer.py → build_pdp_extras.py (hero pass · ?ob=c · package wiring · paint-first tags · ends with the Shopify handoff) → build_pdp_read50.py → tools/qa/mobile_readability.py products/nailscience products/nail-restoration-kit/ (must read 0 at 375/390); checkout = build_checkout_tbyb.py (folds in build_order_summary.py + build_checkout_read50.py); OTOs = build_oto.py. Proofs after any deploy: verify_tracking.py browser --site <base> · verify_tracking.py webhook --site <base> · tools/qa/shopify_matrix.py --quick · tools/qa/nailscience_tbyb_order.py --lane ladder|sub --base <base> --keep-webhook. Still not wired: the 15-pen bundle render (the 5-kit row shows the 9-pen image); the Shopify ad channels (ask ns-shopify-finish). The trial's $6.99 attach is LIVE since 2026-10-15 23:00 PT (ref 2).
| Surface | URL | State |
| Home · PDP | https://trynailscience.com (home, lane C) · https://trynailscience.com/products/nail-restoration-kit/ (PDP) | LIVE — the relaunch build since 2026-10-15 11:20 PT (paint-first tag boot 12:00): Ladder A radio box + the 14-Day Free Trial row, the ?ob=c column, the hero pass, the 50+ pass, every row → our 29next checkout since 2026-10-21 (the Shopify handoff dormant — ?co=shopify previews it; no Buy with Shop Pay); old /?pkg= and /?ob= 301 to the PDP · the Clinicians' Choice card is OFF since 2026-10-21 (the smaller badge = ask ns-clinicians-badge) · the clinicians chip (board D) under the rating since 2026-10-21 afternoon; the offer box rows carry no struck price / Free Shipping line and no trial row |
| Checkout | https://trynailscience.com/checkout/ · staging https://nailscience-staging.netlify.app/checkout/ | LIVE = the Try Before You Buy stamp (2026-10-15): Ladder A rows on pkgs 10/11/2 (books — the ladder dummy order 20/20 on production), the trial card (pkg 12 / 13 via ?pkg=12&plan=tbyb-sub or plan=tbyb-one), the Total Order Summary, SECRET50, 30-Day everywhere, one hashed CSS bundle, no Swiper, no hero bytes on phones. The trial BOOKS since 2026-10-15 23:00 PT on the $6.99 attach (ref 2): monthly + one-kit 26/26 each on staging, monthly 24/24 on production iPhone (order 104892 — $6.99 today, day-14 $24.99 scheduled, renewals ship free); before the attach 29next refused the $0 first order ("Orders with subscription products require payment."). Shipping Protection hidden since May (8ed06a76c). Since 2026-10-21 the ladder rows check out HERE again (pkg 2 / 10 / 11 — the single dummy order 20/20); the Shopify path is dormant (?co=shopify). · since 2026-10-21 a ?pkg= link collapses the selector to the chosen row ("Your Package · Change") · rows mirror the box since 2026-10-21 afternoon: no struck anchor, no Free Shipping line, no trial pick in the full selector · since 2026-10-22: a ?pkg= arrival shows NO package section (the one-row + Change state lasted a day), and the checkout carries NO order summary (David: "remove this") — nothing on the page states the package or the total |
| OTOs | /special-offer/bundle/ · /special-offer/upgrade/ · /special-offer/vip/ | LIVE — VIP = the house text-only page (rebuilt 2026-10-15, pkg 7, $14.99 + $9.99/30d); upgrade (pkg 9, $11.99) + bundle (pkg 8, $29.99) = the May pages + the same-origin shim; chain proven on production (accept / skip / skip → thank-you) |
| Thank-you | /thank-you/ | LIVE — the May page + the shim; ReceiptView only (Pattern D) |
| Legal | /pages/privacy /pages/terms /pages/refunds /pages/shipping /pages/contact /pages/accessibility /pages/medical-disclaimer (pretty paths 301) | LIVE (200) — the one-sided stack for a topical kit, 30-Day Money-Back Guarantee |
| NS-P1 · 5 Reasons | https://trynailscience.com/blog/five-reasons/ | LIVE — mirror of prelanders/nailscience/5reasons-v3 |
| NS-P2 · Top 5 | https://thehealthreportdaily.com/blog/ranking/ (W3 AD024 + W4 AD025's link, /blog/ranking 301s to the slash on the same domain) · copy https://trynailscience.com/blog/ranking/ | LIVE at the April URL since 2026-10-16 (David: "i need the old prelanders re-deployed to the same URLs that were working before") — prelanders/nailscience/top5-advertorial (30-day since 2026-10-14) staged by tools/deploy-health-report.sh; buttons open the PDP in a new tab with the identity forwarded; proof 41/41 |
| NS-P3 · Dr. Franks advertorial (Foot Doctor Reveals) | https://thehealthreportdaily.com/ (W1 AD031 + W2 AD013's link) · copy https://trynailscience.com/blog/foot-doctor/ | LIVE at the April URL since 2026-10-16 — prelanders/drfranks (its nested old Top 5 twin never ships) staged at the hub site's root (the root's 302 to the SkinScience /investigation/ is gone; no SkinScience ad links to the bare root); buttons → the PDP; proof 41/41 |
| Quiz | https://quiz.trynailscience.com | LIVE (2026-05-17 build) |
| Collection | https://trynailscience.com/collection/ | LIVE since 2026-10-15 (200; the handoff on its ladder rows) |
| CAPI relay · order webhook | /api/capi · /api/order-webhook | LIVE — the relaunch functions (SKU gate ORDER_SKU_PREFIX=nailscience, the test-order gate, the TBYB lane map), webhook 67's secret aligned 2026-10-15; probes ALL PASS on production |
| Attribution runtime | /tracking/ns-attribution.js (deferred) + tracking_head.py (MO-Guard → dataLayer → GTM-P9MSJBP7 → pixel init, paint-first) | LIVE — browser proof ALL PASS on production (PDP · checkout · trial · VIP · thank-you) |
| Media plan 2026-10 (the turn-on plan) | https://mediaplan-hq.netlify.app/nailscience/ (schedule board: https://mediaplan-hq.netlify.app/nailscience/schedule/) | Round four ON from 00:00 Oct 31 (campaign 120250114416430510, $50, AD058–AD062); rounds two + three paused by David Oct 30; round one off since Oct 19 (Fotini) |
| Launch Center workspace | https://launch-center-hq.netlify.app/nailscience/ | registered 2026-10-13 |
| Brand DB | https://launch-center-hq.netlify.app/docs/nailscience/brand-db.html | this file, rendered |
| Pages manifest | creative/nailscience/pages-manifest.json | NS-P1 · NS-P2 · NS-CHK · NS-UP1–3 |
- Copy sheet (2026-10-28, David: every headline and primary-text idea judged on one spreadsheet-style grid, on every brand — copy law CL-23): https://copy-sheet-hq.netlify.app/nailscience/ — 28 rows (15 headline, 13 text), nothing judged yet;
python3 tools/angle-studio/copysheet.py build --brand nailscience --auto·verdictsfolds the 👍 / 👎 intobriefs/nailscience/copy/. · since 2026-10-30 the brand's page on Copy Sheet HQ carries every tab: angles · hooks · hook buckets · ad headlines · ad primary text · email subject lines · the truth gate · every verdict list (the Launch Center's grids and card views retired behind redirects)
- Truth Gate (2026-10-29): https://launch-center-hq.netlify.app/docs/nailscience-truth-gate.html · Angle board r1: https://copy-sheet-hq.netlify.app/nailscience/?tab=angles · Hook Studio: https://copy-sheet-hq.netlify.app/nailscience/?tab=hooks · Hook Buckets: https://copy-sheet-hq.netlify.app/nailscience/?tab=buckets · Copy Sheet: https://copy-sheet-hq.netlify.app/nailscience/
- Angle board r2 (2026-11-11): https://copy-sheet-hq.netlify.app/nailscience/?tab=angles (22 angles: r1 NS-A01–A12 GO, r2 NS-A13–A22 awaiting his verdict) · Hooks batch 6 + the weird twenty: https://copy-sheet-hq.netlify.app/nailscience/?tab=hooks · seeds
briefs/nailscience/angles/seeds.md(# Round 2) · ledgerbriefs/nailscience/angles/COPYCHIEF.md(angles round 2 · headlines, hooks batch 6) - Round-2 prelanders (live 2026-11-12, no traffic yet): https://thehealthreportdaily.com/blog/breast-milk/ (NS-P4) · https://thehealthreportdaily.com/blog/grandmas-remedy/ (NS-P5) · https://thehealthreportdaily.com/blog/my-mothers-toenail/ (NS-P6) · https://thehealthreportdaily.com/blog/hoof-test/ (NS-P7, review build until the shoot) · sources
prelanders/nailscience/<slug>/flow.json→build_article.py· deploytools/deploy-health-report.sh
- Media plan round three ON 00:00 Oct 30 → PAUSED by David 09:42 PT Oct 30 (2026-10-29): campaign
120250102841410510(ABO $20, one ad set, AD051–AD057 = the 7 weird r2 keeps) → NS-P2 · planbriefs/nailscience/launch/2026-10/NS-P2-STATICS-R2-ABO.json· the r2 board https://launch-center-hq.netlify.app/docs/nailscience/boards/statics-r2/ (JUDGED: 7 keeps) - Media plan round two ON 00:00 Oct 30 → PAUSED by David 15:38 PT Oct 30 (2026-10-29): campaign
120250100703970510(CBO $50, two ad sets, AD036–AD050) → NS-P2 at its April URL · planbriefs/nailscience/launch/2026-10/NS-P2-STATICS-R1-CBO.json· the media plan board https://launch-center-hq.netlify.app/nailscience/media-plan/ · the weird round two board (on the saved frameworks): https://launch-center-hq.netlify.app/docs/nailscience/boards/statics-r2/ - Media plan round four ON from 00:00 Oct 31 (2026-10-30): campaign
120250114416430510(ABO $50, one ad set120250114416880510, AD058–AD062 = the five r3 keeps) → NS-P2 at its April URL · planbriefs/nailscience/launch/2026-10/NS-P2-STATICS-R3-ABO.json(+ state) · the only campaign of the media plan on from midnight; Fotini's two (§4) run beside it. - Statics r4 on the J&T board — versions of round four's two earners (2026-11-05): http://localhost:8769/dashboard/#nailscience-ads?round=4 (30 cards, 1:1, AD058's and AD061's lines word for word on new pictures; verdicts in
tools/jt/judgement-nailscience-ads-latest.json; round filebriefs/nailscience/statics/r4-concepts.json; images hosted for the launch at /docs/nailscience/boards/statics-r4/img/) · 27 kept, LIVE as round five from 00:00 Nov 6: campaign120250214088960510, AD063–AD089 - Media plan round five — ON 00:00 Nov 6 (2026-11-05): campaign
120250214088960510, CBO $100 a day (raised from $50 before it went on), four ad sets, 27 ads AD063–AD089, the Top 5 page on the proven pair. Round four120250114416430510stays on at $50 beside it. - Videos, LIVE since 2026-11-11 07:37 PT ($50 a day each, the account's only spend): PDP campaign
120250299393720510(ad set120250299393830510) and Top 5 campaign120250299394600510(ad set120250299394650510), AD090 video 01 hook 1 · AD091 hook 3 · AD092 V12 (30-day trim), the proven pair, US 18–65+ Advantage+ audience. Round five120250214088960510paused by David Nov 10. The Top 5 ad set paused by David Nov 11 16:53 PT; the PDP ad set runs on. - Videos R2, PAUSED, ON at 00:00 PT Nov 13 (2026-11-12, $50 a day each): the four animated videos AD093 (02 the fungus) · AD094 (03 the tray) · AD095 (04 the massage) · AD096 (05 the boot), PDP campaign
120250315675400510(ad set120250315675490510"NS-01 PDP · R2 · videos") and Top 5 campaign120250315677630510(ad set120250315677830510"NS-P2 Top 5 · R7 · videos"), the proven pair, R1's targeting; agentcom.davidsanghera.ads-on-20261113-0000. $150 a day on the account from midnight. - Cuts R1, PAUSED, ON at 00:00 PT Nov 13 (2026-11-12, $50 a day): AD097 the toe models · AD098 three mistakes, after summer · AD099 three mistakes, the pedicurist in fall (Lunavia's openings, our close), PDP campaign
120250317776060510(ad set120250317777940510"NS-01 PDP · cuts R1 · videos"), the proven pair, R2's targeting; the switch rides agentcom.davidsanghera.ads-on-20261113-0000with R2. Record:tools/jt/one-offs/nailscience-cuts-2026-11/. - R2 + cuts R1 ran Nov 13–15 and are OFF (read 2026-11-15 22:40 PT). All three went ACTIVE at 00:00 PT Nov 13 (
~/Library/Logs/ads-schedule-on.log), and David's account paused them on Nov 15: both R2 campaigns at 09:26 PT, cuts R1 at 20:39 PT. Lifetime numbers: R2 PDP $185.15 · 1 purchase · 8 checkouts · 34 LPV; R2 Top 5 $174.17 · 0 · 7 · 47; cuts R1 $199.47 · 2 · 14 · 70. All three were over the bars ($43 an order or $4 a checkout; off at $60 with no order). By code (ad_codes.py report --brand nailscience --pages --since 2026-11-13): **AD097 the toe models took all of cuts R1's spend, with a 9.41% CTR (the round's best), 70 LPV and 2 buys at $99.73** (ROAS 0.90); AD098 and AD099 never delivered. AD094 the tray spent $166 for 0 buys, AD093 the fungus $101.54 for 1 buy (PDP), AD095 $65.66 for 0, AD096 $25.98 for 0. - Statics r3 board — versions of the winners (2026-10-30): https://launch-center-hq.netlify.app/docs/nailscience/boards/statics-r3/ (28 cards, 1:1, anchor = AD025; the round file
briefs/nailscience/statics/r3-concepts.jsoncarries the winners ledger; the QA sheetbriefs/nailscience/statics/wild/2026-10-30-a.sheet-r3.jpg). Round three's campaign120250102841410510PAUSED by David 2026-10-30 09:42 PT; round two120250100703970510LIVE at $50 CBO. - Wild statics r1 board (2026-10-29): https://launch-center-hq.netlify.app/docs/nailscience/boards/statics-r1/ (JUDGED 2026-10-29: 15 kills, 15 launched) · the QA sheet:
briefs/nailscience/statics/wild/2026-10-29-a.sheet-r1.jpg· the cards:creative/nailscience-ads/r1/
9 · Open questions / asks (console lane nailscience, Needs You tab)
- OPEN 2026-11-15: the next launch's strategy. David decides it before any plan is written (his rule, 2026-11-15: "next time we launch, we HAVE To decide the launch setup/strategy";
skills/launch-strategy/SKILL.md). The choice: House 50 (the shape R2 and cuts R1 ran: 3 videos in a $50 ad set, Advantage+ audience on), Solo 250 (DreamScene Pro's setup: each ad its own campaign at $250 a day, a bid cap he names, 7-day click, Advantage+ audience off, Facebook + Instagram) or Solo 100 (added 2026-11-16, his words: "Lets add Solo $100 with a costcap as a method": the same one-ad-per-campaign shape at $100 a day on a cost cap he names, kept on). Three ads cost $50, $750 or $300 a day. AD097 (the toe models: a 9.41% CTR, all of cuts R1's spend) is the obvious first ad under either Solo strategy; both give AD098 / AD099 (no spend in cuts R1) a budget of their own, Solo 100 at $100 a day instead of $250. The caps are his numbers: for Solo 250 the book's starting point is about 2.5× the $43 bar, about $105; for Solo 100 the useful range is about $43 (tight: Meta may hold the budget back) to $100 (above it, one sale a day is already under the cap). - ANSWERED 2026-11-12 (13:50 PT) by his word, "do the cut versions for tonight": three of the five cut and launched PAUSED, on at 00:00 PT Nov 13 (§6, §8: AD097 the toe models, AD098 / AD099 three mistakes, PDP, $50). OPEN from it: (1) Lunavia's own per-ad spend and purchases for these videos (a partner share of its ad account or an export; our token can't reach it); (2) his
SHOTLIST-06/08verdicts, so the day-3 winner is remade in our words; (3) the toe models' "before the holidays" opening and "28 years I did feet for a living", flagged, his word pulls AD097 before midnight. Was: OPEN 13:20 / 13:40, five Lunavia Ad Library videos sent to launch ("Launch these into another $50 campaign for tomorrow", then "Lets launch them for $50/day spend so we can see which one works and which one to remake"); "Lunavia is MY BRAND also"; held from a straight launch for the offer mismatch. Not cut:1095551209965205(the polish reply, the offer from its first line) and1149101697535845(the drawn story, bees + propolis). - OPEN 2026-11-12 (night): his look at the second cuts of 11 (the problem first, on his note) and 12 (QA fixes) (§6), both sent; then 06 to 10 as they finish.
- OPEN 2026-11-12 (afternoon), all seven sent by 2026-11-13 03:20 PT: his look and his ear on the made remakes (§6): second cuts of 06, 07, 08, 10 and 11, the fourth of 12, the second of 09, each with its
QA-NN.md, masters increative/nailscience/videos/<NN-slug>/finish/. Nobody has heard the voices. The calls still built as written: the doctor in 09, who speaks (generated presenters with no profession), a price, the customer count, 10's small-print line, the salon line in 08 and 09. Where to give it (2026-11-13 evening): every NailScience video (01 with its four hooks, 02 to 12) is on Video Review HQ, https://video-review-hq.netlify.app/ : select the part you mean and leave the note there (typed, a reason, a drawing on the frame, or a voice note); a remake plays beside its original.tools/video-review/pull.py nailscienceturns the notes intovideo-studio/scripts/nailscience/NOTES-NN.md. - PARTLY ANSWERED 2026-11-12 (afternoon) by "generate video 11 in the animated lane, rest lets render them": who makes each (11 the studio, the rest generated by us). Was: OPEN 2026-11-12 (midday): his verdicts on seven shot lists,
SHOTLIST-06to12(§6), and the calls at the foot of each.** Nothing is filmed or generated before them. The calls that change the most: (1) who speaks in 06, 07, 08, 10 and 12: a generated presenter with no profession, as written, or a real person (a nail technician can say her own years; nobody is written one); (2) the doctor character in 09 (55 seconds, no name): yes or no; (3) the customer count the references show and ours do not (s1, not ruled on; no customer clips on file); (4) a price in the offer (none is said, except "less than that bottle of oil" in 10); (5) 10's small-print line, and whether our own Step 2 box carries "not effective on the nail"; (6) the salon line in 08 and 09 (looked up, not through the Truth Gate); (7) who makes each: 06, 07, 08, 10 and 12 the editors · 09 a generated film or six actors · 11 the animated lane, like video 02. Also his: the seven sets of five hooks onto the Copy Sheet (CL-23), and the Apify token (refused, 401). - ANSWERED 2026-11-12 (midday) by his word, "Lets remake each (1-7) on close to verbatim in the pacing, scripting, shot list, and beats": all seven are written as scripts 06 to 12 (§6). Was: OPEN 2026-11-12 (morning): which of Lunavia's videos to re-make (§6). The pick list is
ads/research/nail-fungus/pages/shannon-fitzgerald/RECREATE-2026-11-12.md: (1) the toe models, (2) four things you should never do, (3) three mistakes, (4) the bride's mother, (5) my receipts, (6) the three villains, (7) "Hey Jessica". Mine: 1 first, then 2. On his pick: our own script on that structure from NS-A01 (hooks, the chief, the read-aloud), then the shot list for the editors. Also his: renew the Apify token (refused, 401), or leave the sweeps on the free lane. - ANSWERED 2026-11-12 (~09:15 PT) by his launch word, "Launch Video 02,03,04,05 for NailScience $50/day using the prelander and PDP": all four kept and launched as made (§6, R2 on at 00:00 PT Nov 13). OPEN from it: "the prelander" was read as the Top 5 page (R1's leg, whose ad set he paused Nov 11 16:53 PT); a swap to one of the four new prelanders before midnight is his word away. Still open below: the other hooks, his next pick, his 15 clips, the StefanBrain request. Was: OPEN 2026-11-12, the four animated videos are on the J&T board (§6): his verdicts.
ns-v02-fungus·ns-v03-tray·ns-v04-massage·ns-v05-boot. His to rule inside them: (1) the voices, all four (nobody has listened; 02's are the studio's own); (2) 02's three words that differ from the script ("Your cream", "I'm packing", an added "Ah, oof"); (3) 03's pens with the brand's name on them; (4) 04's one shot of her face. Then: which go to Meta, on his word (nothing is launched). Not made: the four other hooks of each (one new opening shot each). Still open from the night before: his next pick fromCONCEPTS-r3.md(N5 is now animated; the hoof as an animated illustration) · A17, A19, A20 are GO in the bank only · his 15 studio clips (Video 1 to 15) · a request to StefanBrain's team for a video tool on the MCP, a message in his name, so on his word. - ANSWERED 2026-11-12 (00:00 PT), was OPEN 2026-11-11 (late night): a yes on four shot lists (§6). "Yes, go ahead and generate all 4 videos. N5 should be animated. #1 yes #2 good #3 all good, keep going, forget the Truth Gate. #4. You can generate these." As asked: (1)
SHOTLIST-02to05: keep, kill or tweak per shot. Nothing is generated before that. (2) His next pick fromCONCEPTS-r3.md, if he wants more than the four. (3) The hoof (N1): whether he or a creator films it. (4) Three research items for the Truth Gate: the nail can crumble and lift, the clinic softens first, sweaty shoes as a risk factor. (5) A17, A19 and A20 are GO in the bank only; the board still shows them unjudged. (6) His 15 studio clips (Video 1 to 15): on the board, into a cut, or left in the studio. (7) A request to StefanBrain's team for a video tool on the MCP: a message in his name, so on his word. ANSWERED the same night: the first animated video is the fungus ("1. Yes"). - OPEN 2026-11-11: the post-ID launch, $50 a day (§6). The post is back but cannot run as an ad (a purple-background text post; Facebook: not eligible). Asked him: (1) the purchase version (recommended): the same purple card as the picture of a new Carol Anne Foster ad, a Learn More button to the PDP (where her UPDATE points), Purchase-optimized, $50 a day from 00:00 PT; it starts at zero reactions and her UPDATE + before/after can be posted and pinned on it by hand in Business Suite; or (2) keep the 33 reactions / 32 comments: he edits the post to add a photo (that drops the background) and it may become promotable — then likes-and-comments optimization only (the link is in the comments), after adscaler gains an existing-post creative. Truth Gate on the card's words stands ("celebrities are using").
- OFFERED 2026-11-11 (§6, the uncredited sale): two attribution fixes, on his word. (1) Send the VIP / add-on value to Meta (today the Purchase carries the kit only); (2) a phone field at checkout plus the webhook's
phone_numberpath, so a buyer with a blocked browser can still be matched. Neither recovers order 106403 (no click ID exists to send). - OPEN 2026-11-11: angle board r2 (§6) — GO / ADAPT / PARK / KILL on NS-A13 to NS-A22 (ask
ns-angle-board-r2). A GO routes the angle's packets (angles.py route --brand nailscience --angle <ID> --to all) for the next statics and video rounds; a KILL becomes a standing ban (CL-18). Also open: the refunds page's CellScience paragraph and the PDP's 90-day guarantee block (a separate task, not changed). - OPEN 2026-11-12: the four round-2 prelanders are live for his read (§6, §8) — his notes rewrite
flow.jsonand rebuild; the photographer's photos replace the generated ones; NS-P7's results are written only from the real test; then which pages take traffic (a media-plan round). Also open: NS-P3's 9–18 months / 95% / Trustpilot / sticky-forwarding lines (seen, not changed). OPEN: quiz concepts r1 — which of the five (plus r2's #10) to build (briefs/nailscience/prelanders/QUIZ-CONCEPTS-r1.md). - ANSWERED 2026-11-11 on the board + in chat: hooks 1 and 3 kept, hook 2 killed, launched with the rap to the PDP and the Top 5 page (§6). Still open from this item: hooks 4 and 5 unrated; the 53 s vs 40 s cut; shot 11; the disclaimer (officer); music. The original ask, his verdicts on video 01's five cuts (§6) — on the J&T board, Group VIDEO (
ns-v01-h1-castthe lead,-h2-pen,-h3-shelf,-h4-year,-h5-shower). (1) Keep / kill / refine per cut; a changed shot is one still or clip remade and a re-cut. (2) Hook 5 is held until the kit's printed directions say when to apply (after a shower or not). (3) The length: 53 seconds as cut, or the 40-second re-cut (loses the "products are" line, the strength line and the clock; no new generation). (4) Whether shot 11 goes on to clear the fungus as the product page's loop does (one clip). (5) A disclaimer on the video: none on file, the officer's call. (6) Music: none; a bed needs a licence. (7) Where it runs: the three-video ad set with the rap (V12), the Top 5 page, the proven headline and text, created paused, on at midnight, through adscaler on his word. Still open from round 2: the pharmacist line and the news desk (second in the order), whether the page's customer media is cleared, the rival-video board. Closed here: which pens are in a kit (three: one softener, two antifungal). - ANSWERED 2026-11-10 (night), in chat: the pick = "go with concept 1 first" (§6). The round-2 ask below is otherwise still open. (1) His pick: concept 1 (the faceless explainer on the shield) first and the news desk second, or another order; both run with the rap. (2) The pharmacist line, when the news desk is made. (3) Where the videos land (default the Top 5 page). (4) Whether the product page's customer videos and before-and-afters are real and cleared for ads. (5) Whether he wants the rival videos on a Launch Center board to watch and mark, and the angle board refreshed on Copy Sheet HQ (both need a deploy; the October natives board is still unjudged: 81 ads). The round-1 version of this ask, below, is superseded.
- ~~OPEN 2026-11-10 (asked in chat, where he asked for the pitch): the video concepts (§6).~~ Superseded the same evening by round 2 above. (1) His pick: one concept, or two to fill the three-video ad set with the rap. (2) The pharmacist line ("Pharmacist tested 7 toenail kits, only 1 actually works"): voiced in video as it runs on the statics, or held until the Truth Gate's s5 is ruled. (3) Where the video lands: the Top 5 page (the default, every winner runs there) or the product page. Not blocking the pick, needed before a script: the kit's printed directions on when to apply · a source for shoes and bath mats carrying the fungus · whether real clippings and a real hoof shoot are possible. Nothing renders before his yes and a signed-off shot list.
- ANSWERED 2026-11-05: round five's budget and round four —
ns-r5-budget, "Do $100/day" + "Yes current one stays on" (§6).ns-jt-r4-judge, 27 keeps + "Okay launch all the kept ones". - ANSWERED 2026-11-02:
cs-order-summary→ B (David in chat) — the soft card shipped to every checkout but Sculptape the same night (§6) - OPEN 2026-10-26: phone at checkout (§6) — the one attribution lever left; levers 2 + 3 SHIPPED the same day ("do 2 and 3").
- CLOSED 2026-10-26: the QA-order leak (§6) — runner blocks tracking, webhook reads the full order; proven with two test orders. (Was: OPEN 2026-10-22: Meta purchase counts include QA orders (§6)) — read round one off real orders (6 since Oct 20), not Ads Manager (17); fixes proposed: the order runner blocks tracking, the webhook reads the full order (test flag, subscriptions, partner_sku).
- CLOSED 2026-10-21 afternoon:
ns-clinicians-badge→ D, the avatar chip (David in chat), shipped. The 10-16 list below still stands. - ~~OPEN NOW (2026-10-21):
ns-clinicians-badge(choice A–E or none — the smaller clinicians' badge; board/docs/nailscience/clinicians-badge/board.html)~~ - OPEN NOW (2026-10-16):
ns-shopify-finish(re-posted: Facebook & Instagram done by David; Google & YouTube left) ·ns-ads-go(re-posted: everything it waited on is ready — name the day).ns-ads-destinationANSWERED in chat 2026-10-16 = re-deploy the pages at the same URLs (done, §6).ns-webhook-crossfire: David "Ok" to the explanation (withdrawn 10-15). - OPEN AS OF (2026-10-15 night):
ns-shopify-finish(re-posted: the time zone is done, the two ad channels are not on the store) ·ns-ads-destination(David asked what it means — explained in chat; waits on his yes) ·ns-ads-go(the day + $150/day; nothing switches on without it). Everything below is the history of this list. - OPEN (as of 2026-10-14):
ns-webhook-crossfire(since 2026-10-13) ·ns-shopify-store(the owner's five Shopify steps, posted 2026-10-14) · media plan 2026-10 (posted 2026-10-14 evening):ns-ads-destination(the old domain 301s to the brand domain; the ads stay untouched) ·ns-video-cut(V12's 90-day line vs the 30-Day lock, and its page) ·ns-ads-go(the day + $150/day; nothing switches on without it). ·ns-campaign-reactivate(the 29next campaign is off; the trial lane waits) ·ns-staging-review(judge the offer box / homepage / footer / legal on staging) ·ns-shipping-promise(Ohio same-day vs Shenzhen 7–14 days — which is true). - ANSWERED 2026-10-14:
ns-relaunch-base·ns-checkout-platform·ns-offer·ns-formula-truth·ns-inventory·ns-accounts·ns-guarantee(on the hub) ·ns-offer-pick= Ladder A (in chat) — verdicts in §6. - 2026-10-14 (late):
ns-tbyb-shippingposted (the $6.99 trial shipping method — one admin screen, then I attach it and re-stamp) ·ns-campaign-reactivatewithdrawn (the campaign answers again) ·ns-webhook-crossfireis answered in practice byORDER_SKU_PREFIX=nailscienceon the production env (live the moment the relaunch build deploys to production) — the ask stays open for his word. - 2026-10-15:
ns-shopify-finishposted (the store time zone + the two ad channels) ·ns-shopify-storewithdrawn (the store, Shopify Payments, our app and the DNS are done). - 2026-10-15 (midday) — the site is LIVE:
ns-tbyb-shippingRE-POSTED with the gateway's refusal (a $0 first order carrying a subscription is refused — the free trial cannot take an order until the $6.99 line exists) ·ns-staging-reviewstays open as the live-site review (judge trynailscience.com now) ·ns-webhook-crossfireanswered in practice since 11:20 (ORDER_SKU_PREFIX=nailsciencelive; a foreign SKU is ignored — proven) ·ns-shopify-finishstill open (Shopify orders report to no ad channel until the two channels are on) ·ns-shipping-promisestill open.
- 2026-10-15 (night) — David answered in chat ("#1 You do it through my browsers. #2 Its done. #4 what does this mean? #5 thats okay. #6 its okay #7 all good. #8 what does this mean"):
ns-tbyb-shippingANSWERED (done in his Chrome — §3, §6) ·ns-video-cutANSWERED = run it as is on the Top 5 page ·ns-shipping-promiseANSWERED = keep the page as it is ·ns-staging-reviewANSWERED = yes (all recorded on the hub with his words, who "David (chat)") ·ns-webhook-crossfireWITHDRAWN (resolved: the SKU gate live since 11:20; explained in chat) ·ns-shopify-finishRE-POSTED (time zone verified America/Los_Angeles; the Facebook & Instagram and Google & YouTube channels are NOT on the store — publications = Online Store · Point of Sale · Shop, and the storefront + checkout fire no Meta or Google request; neither of his Chromes is signed into this store) ·ns-ads-destinationexplained, open ·ns-ads-goopen (not answered).
- ANSWERED 2026-10-30 ~15:39 PT:
ns-statics-r3-judge= five keeps on the board + "okay lets relaunch in a campaign for $50" → round four (§6).ns-truth-gatestill open. - OPEN 2026-10-29:
ns-truth-gate(53 claims). ANSWERED in chat 2026-10-29 ~21:45 PT:ns-statics-r2-judge= 7 keeps on the board, "i kept some, we can launch a $20 ad for those next" · "starting midnight tonight" → round three launched (console update posted). ANSWERED ~17:00 PT:ns-statics-r1-judge= 15 kills on the board, "rest launch for nailscience for $50/day with the best headlines and prelander" → the 15 launched (console update posted);ns-third-slot= answered by the same word (the third $50 is round two's campaign).ns-angle-board-r1= GO on all twelve; the statics gates = all kept, all thirty rendered.
10 · Change log
- 2026-11-16 — Third cuts of 07 and 09 on his Video Review HQ notes (§6):
07.shots.json(stills spores-a and spores-b, clip spores-spread, shots s13b and s13c, a 4:5 caption rule),07.remake.json(row 13, differs),SHOTLIST-07.md,07.finish.json,QA-07.md;09.shots.json(stills s1-*, talk t0101–t0105, s001's 4:5 crop, s004's push-in),09.remake.json(S1, rows 1, 4 and 5, differs),SHOTLIST-09.md,09_closed-toe-heels-on-the-sand.md(beat 1's picture note, Visual style),09.finish.json,QA-09.md, the scripts README. Masters and review copies are not in git. - 2026-11-16 (morning) — Solo 100 in the launch book (§9; David: "Lets add Solo $100 with a costcap as a method"): one ad per campaign at $100 a day on a cost cap he names (
skills/launch-strategy/SKILL.md,tools/launch_strategies.py); Meta accepted a $100 ad set on a cost cap at $43 to $150 in a validate-only check on act_832705320684189 (nothing created). - 2026-11-15 (22:40 PT) — The Nov 13 round read and the launch-strategy rule (§8, §9): R2 + cuts R1 lifetime numbers, all three off since Nov 15 (David's account); the next launch picks House 50 or Solo 250 first (
skills/launch-strategy/SKILL.md,tools/launch_strategies.py;tools/ads_launch.pyrefuses a plan without his words). - 2026-11-14 (afternoon) — The AI line off every video (§6):
06to10and12.shots.json(nofinish.disclaimer), their.finish.jsonand06-4x5/12-4x5.finish.json(no disclaimer text), the Disclaimer rows of scripts 06 to 12,12.remake.json,SHOTLIST-12.md, the scripts README; the rule inCLAUDE.mdandskills/product-launch/copy-laws.md(CL-00),skills/video-finish/SKILL.md,skills/video-ad-scripts/SKILL.md§12. Masters and review copies are not in git. - 2026-11-13 (evening) — Every NailScience video on Video Review HQ (§9):
tools/video-review/(publish, build, pull, the engine, three layouts for his pick),tools/deploy-video-review-hq.py; review copies increative/nailscience/videos/<NN-slug>/review/(not in git). - 2026-11-13 (03:20 PT) — 09's second cut on QA-09 (§6, §9):
09.shots.json,09.finish.json,QA-09.md, README. - 2026-11-13 (small hours) — 06 to 10 made; new cuts of 06, 07, 08, 10 and 12 on the QA agent's reports (§6, §9):
06.shots.json,06.finish.json,06-4x5.finish.json,07,08,09and10.shots.jsonand.finish.json,12.shots.json,12.finish.json,12-4x5.finish.json,06.remake.json,07.remake.json,SHOTLIST-06.md,SHOTLIST-07.md,QA-06.md,QA-07.md,QA-08.md,QA-10.md, README;tools/video-shots.py(the join exact in picture and sound, the pitch check, a 4:5 crop instead of a squash,hold_last_caption),tools/video-finish.py(captions.hold_last),skills/video-ad-scripts/SKILL.md(six traps). - 2026-11-12 (night) — 11 opens on the problem (his note); second cuts of 11 and 12 (§6, §9):
11_three-on-top.md,11.remake.json(pre),SHOTLIST-11.md,11.shots.json,11.finish.json,12.shots.json,12.finish.json,12-4x5.finish.json,12.remake.json,SHOTLIST-12.md,COPYCHIEF-06-12.md,QA-11.md,QA-12.md;tools/video-remake.py(an opening of our own:pre),tools/video-shots.py(4:5 composed on its own, frame-exact joins, unheard words re-heard in their gap,finish.colors, face check on for presenters). - 2026-11-12 (afternoon) — The remakes made: 11 and 12 built, 06 to 10 in progress, the QA agent (§6, §9):
video-studio/scripts/nailscience/11.shots.json,11.finish.json,12.shots.json,12.finish.json,SHOTLIST-11.md(as made),tools/video-shots.py,tools/video-remake.py(qa),tools/video-finish.py(label colours),tools/person-matte.swift,.claude/agents/video-remake-qa.md,.claude/commands/video-remake-qa.md,skills/video-ad-scripts/SKILL.md§13; media increative/nailscience/videos/(not in git). - 2026-11-12 (13:55 PT) — Lunavia cuts R1 launched PAUSED, on at 00:00 PT Nov 13 (§6, §8, §9): plan + state
briefs/nailscience/launch/2026-11/NS-01-CUTS-R1.json,briefs/nailscience/ad-codes.json(AD097–AD099), one-offtools/jt/one-offs/nailscience-cuts-2026-11/(uploads, thumbnails, transcripts, README). - 2026-11-12 (13:20 PT, updated 13:40) — Five Lunavia ads sent to launch, held; Lunavia is David's brand too (his word) (§9): this file only; nothing on Meta, no ad codes.
- 2026-11-12 (midday) — The seven remakes, scripts 06 to 12 (§6, §7, §9):
video-studio/scripts/nailscience/(seven scripts, sevenNN.remake.json,SHOTLIST-06to12,COPYCHIEF-06-12.md, README),email/nailscience/video-scripts-audio/(06 to 12, mp3 + json),ads/research/nail-fungus/pages/shannon-fitzgerald/(shots/for seven cuts, V02 re-transcribed,RECREATE-2026-11-12.mdcorrected and extended),tools/video-remake.py(new),tools/angle-studio/rival_page.py(shots),tools/audit-video-scripts.py(Pace,spoken_name),tools/read-aloud-scripts.py(Voices),skills/video-ad-scripts/SKILL.md§12,skills/angle-studio/SKILL.md. - 2026-11-12 — The four animated videos launched PAUSED, on at 00:00 PT Nov 13, PDP + Top 5, $50 a day each (§6, §8, §9): plans + state
briefs/nailscience/launch/2026-11/NS-01-VIDEOS-R2.jsonandNS-P2-VIDEOS-R2.json,briefs/nailscience/ad-codes.json(AD093–AD096), one-offtools/jt/one-offs/nailscience-videos02-05-2026-11/(uploads, thumbnails, the four transcripts, README); David's Nov 11 pause of R1's Top 5 ad set recorded from the activity log. - 2026-11-12 (morning) — A rival's whole page read: Lunavia (§6, §7, §9):
ads/research/nail-fungus/pages/{shannon-fitzgerald, lunavia-network}/(the sweeps,page.json, transcripts, posters,RIVAL-PAGE-2026-11-12.md,RECREATE-2026-11-12.md,notes.json),tools/angle-studio/rival_page.py(new),tools/fb-ad-library-page-sweep.py,tools/angle-studio/rival_videos.py,skills/angle-studio/SKILL.md. - 2026-11-12 — Round-2 prelanders built and deployed (§6, §8, §9):
prelanders/nailscience/{build_article.py, breast-milk, grandmas-remedy, my-mothers-toenail, hoof-test, _photos, COPYCHIEF-r2-pages.md},tools/deploy-health-report.sh,creative/nailscience/pages-manifest.json(NS-P4–P7 live_url). - 2026-11-12 — The four animated videos made and on the board (§6, §7, §9):
video-studio/scripts/nailscience/gains02.shots.jsonto05.shots.json,02.finish.jsonto05.finish.json, a Made section inSHOTLIST-02to05, the as-made note in script 02;README.mdandCONCEPTS-r3.mdupdated;tools/video-shots.py(scenes,splice,voice --from-file,smooth, labeluntil/off_at, the early-caption fix) andtools/video-finish.py(anchors);skills/video-ad-scripts/SKILL.md§11 (one film per video, what the studio gets wrong and how it is caught). Media increative/nailscience/videos/(not in git). Nothing deployed, nothing on Meta. - 2026-11-11 (late night) — Video concepts round 3, four picks scripted (§6, §7, §9):
video-studio/scripts/nailscience/gainsCONCEPTS-r3.md(ten animated, ten not, the weird twenty), scripts02_to05_(round 2),COPYCHIEF-02-05.md,SHOTLIST-02to05; read-aloudsemail/nailscience/video-scripts-audio/02to05; the bank: NS-A17, A19, A20 GO for video on his chat word (local fold), their video packets; the studio test clip (Video 16) filed. - 2026-11-11 — His picks (1, 2, 7, 9), the 3-month clock, quiz concepts r1 (§6, §9): truth-gate.md + truth-gate-claims.json (b3 rewritten, p9 withdrawn), seeds.md (p9 out), CONCEPTS-r2.md, QUIZ-CONCEPTS-r1.md.
- 2026-11-11 — Prelander concepts r2 (§6, §9):
briefs/nailscience/prelanders/CONCEPTS-r2.md(his breast-milk line as the model, #9 kept, eight rebuilt from the buckets), r1 marked with his verdict; nothing drafted. - 2026-11-11 (night) — StefanBrain's video studio wired in as the animated lane (§6, §7, §9):
tools/video-shots.pygainsrenders·take·dropand thestefanbrainengine name (an app-made clip is filed, never called;motionleaves it alone);skills/video-ad-scripts/SKILL.md§11 "The animated lane" (what was checked, the studio, the route, the fences); David's words in §6; the ask in §9. - 2026-11-11 — Prelander concepts r1 (§6, §9):
briefs/nailscience/prelanders/CONCEPTS-r1.md, ten concepts on bank angles and buckets, nothing drafted. - 2026-11-11 — Angle Studio round 2 (§6, §8, §9): ten seeds NS-A13–A22 (
angles/seeds.md), 70 hooks in batch 6 + the weird twenty (angles/hooks.md), the chief ledger (angles/COPYCHIEF.md), bank + ANGLES.md distilled, trending refreshed, boards rebuilt (round 2, 54 new rival thumbs undertools/brand-hub/docs/nailscience/angles/), Copy Sheet HQ deployed, askns-angle-board-r2posted. - 2026-11-11 — Why order 106403 (AD090, PDP) is not on Meta (§6, §9): a browser that blocked Facebook's tracking (no fbclid, no
_fbp/_fbc); Meta received the server Purchase without the click ID; the PDP ad path re-proven in a beacon-blocked browser; the add-on value gap and the phone path recorded, nothing changed. - 2026-11-11 — The post-ID launch: the post is back (a purple-background text post, 33 reactions, 32 comments, the author's UPDATE to the PDP) and still not promotable; the ask revised to the purchase version vs editing the post to add a photo (§6, §9).
- 2026-11-11 — The post-ID launch asked for and not launched: the named post is deleted; today's question post is text only and not promotable; how a post ad works, why a link-in-the-comments post cannot optimize for purchases or any 'micro' website step, and the three shapes (§6, §9).
- 2026-11-11 — The three kept videos live on Meta, PDP + Top 5, $50 a day each (§6, §8, §9): plans + state
briefs/nailscience/launch/2026-11/NS-01-VIDEOS-R1.jsonandNS-P2-VIDEOS-R1.json,briefs/nailscience/ad-codes.json(AD090–AD092), one-offtools/jt/one-offs/nailscience-video01-2026-11/(uploads, thumbnails, the rap's transcripts before and after the trim, README). - 2026-11-10 (late night) — Video 01 made on David's "Yes, please use the best models to generate them" (§6, §7, §9):
tools/video-shots.py(new: stills · photos · voice · motion · cut · finish · sheet · lock · cost, hook variants, a cost ledger),tools/video-finish.py(thenailsciencekit, a labels layer placed on the 9:16 source so both masters put a label over the same spot, a text + sub line on the close pill,"facecheck": falsefor a cut with no face),video-studio/fonts/nailscience/DMSans-600.ttf+-700.ttf(the page's woff2, converted),video-studio/scripts/nailscience/01.shots.json+01.finish.json+01-h2…h5.finish.json(new),SHOTLIST-01.md(his yes + Made),README.md,01_everything-you-tried-sat-on-top.md(two notes),skills/video-ad-scripts/SKILL.md§11 (the build and what it taught). Media outside git:creative/nailscience/videos/01-sat-on-top/and the five board dropscreative/nailscience/videos/ns-v01-*.mp4+ sidecars, in the worktree and the main checkout. Nothing deployed; nothing on Meta. - 2026-11-10 (night) — Script 01 + shot list on David's "go with concept 1 first" (§6, §9):
video-studio/scripts/nailscience/—01_everything-you-tried-sat-on-top.md(round 2),SHOTLIST-01.md,COPYCHIEF-01.md,README.md,_v1/(round 1); audioemail/nailscience/video-scripts-audio/01.mp3+01.json(+v1/);tools/audit-video-scripts.py(thenailscienceblock),tools/read-aloud-scripts.py(the Duration row carries the day of the render),skills/video-ad-scripts/SKILL.md§11 + two traps. Nothing generated, nothing deployed. - 2026-11-10 (evening) — Rival video research into the Angle Studio; video concepts round 2 (§6, §9):
ads/research/nail-fungus/(the video sweep + raw,video/manifest.json·notes.json·rival-videos.json· 150 posters,scripts/90 transcripts,RIVAL-VIDEOS-2026-11-10.md; the mp4s gitignored),briefs/nailscience/angles/seeds.md(refs + signal on all twelve),bank.json+ANGLES.mdre-distilled (146 creatives, 88 transcripts; verdicts and hooks unchanged),angles/NS-A01…12/video.md(new),video-studio/scripts/nailscience/CONCEPTS-r2.md(new; r1 marked superseded),tools/angle-studio/rival_videos.py(new),tools/angle-studio/angles.py(round-numbered manifest fallback),skills/angle-studio/SKILL.md§3 + §3c + two traps,skills/video-ad-scripts/SKILL.md§0. No board rebuilt, nothing deployed, nothing generated. - 2026-11-10 — Video concepts round 1 pitched (§6, §9):
video-studio/scripts/nailscience/CONCEPTS-r1.md(new: the ask, the sources, the claim fence, what the account knows, ten concepts, the recommendation, the asks, the weird twenty, the steps on his yes). Nothing scripted, boarded or generated; the hook bank is untouched (the openers used are existing chief-PASS rows). - 2026-11-05 (14:47 PT) — Round five at $100 a day; round four stays on (§6, §8, §9): twin v3.6 (round five, its ads, the launch log, the cap note, the two decisions closed), plan doc, the plan JSON's budget, lane answer
ns-r5-budget, Media Plan HQ + Launch Center;adscaler/app/api/campaigns/budget/route.ts,updateCampaignBudgetinadscaler/lib/facebook.ts,adscaler/__tests__/lib/facebook-budget.test.ts. - 2026-11-05 (10:40 PT) — Round five launched paused, on at 00:00 Nov 6 (§6): plan + state
launch/2026-11/NS-P2-STATICS-R4-CBO.json(the unused draft removed),ad-codes.jsonAD063–AD089, round file verdicts, twin v3.5 (round five + its ads + the launch log + two decisions), plan doc, copy sheet, lane answer, Media Plan HQ + Launch Center. - 2026-11-05 (10:20 PT) — Round 4 repointed at the J&T board (§6): BRAND §8/§9, twin + plan doc, launch draft note, round file note, ask
ns-statics-r4-judgewithdrawn forns-jt-r4-judge.creative/dashboard/app.js:#<board>?round=Nopens a board on one round (app.js v123). - 2026-11-05 — Round four read; statics r4 built and boarded (§6): run + shots + sheet
briefs/nailscience/statics/wild/2026-11-05-a*, round filestatics/r4-concepts.json, cardscreative/nailscience-ads/r4/(+ the J&T manifest), boardtools/brand-hub/docs/nailscience/boards/statics-r4/, launch draftlaunch/2026-11/NS-P2-STATICS-R4-ABO.draft.json, twin v3.4 (round four's read, round five waiting on the keeps, the old rounds five and six moved down one), askns-statics-r4-judge.tools/wild-statics/wild.py: a run'sversions_oflets the winning lines repeat on purpose (skill note).tools/brand-hub/statics_board.py: square cards no longer cropped to 4:5, the intro comes from the round file (it still carried CellScience's wording on r1–r3). Nothing launched, nothing changed on Meta. - 2026-11-03 — Media Plan HQ:
tools/mediaplan-hq/(model.py · template.html · build.py · netlify.toml + _headers/_redirects) +tools/deploy-mediaplan-hq.py(gated: freshness, hygiene --internal, no other brand on a plan page, smoke); https://mediaplan-hq.netlify.app/nailscience/media-plan-2026-10/ is this plan, /nailscience/schedule/ its board (media_plan_board.py still writes it into docs/nailscience/media-plan/). Launch Center: /docs/nailscience-media-plan-2026-10.html and /docs/nailscience/media-plan/board.html → 301 to the HQ (build.py write_redirects); the Copy Sheet HQ Media plan tab and mediaplan.py removed. - 2026-11-02 — the order line under the checkout button (§6, David "B"):
tools/checkout_order_line.py+tools/checkout_selector.py(on by default) +tools/qa/checkout_order_line_check.py; NailScience checkout rebuilt (diff = the card), staging → QA → production (7e128dacc, 2ede3d642, 60790a446); the same card on CellScience, SkinScience ×5, HumanFrame ×2, PeptideScience, Charm (production) and Sculptape (committed, deploy pending from /dev). - 2026-11-02 — the order-summary mock-ups (§6, §9): board + 16 captures
tools/brand-hub/docs/nailscience/checkout-summary/, askcs-order-summary; no site change. - 2026-11-01 (21:20 PT) — The orders read (David: "apply the new nailscience orders to the ads"): no Nov 1 order, every order since Oct 26 registered in Meta, the last two on Fotini's Testing CBO; her 01:18 PT cut ($150 → $30, BFCM off); round four after two days $135.61 · 3 checkouts · 0 orders (§4, §6).
- 2026-10-30 (15:45 PT) — Round four LAUNCHED (the five r3 keeps, ABO $50, AD058–AD062, ACTIVE 00:00 Oct 31); rounds two and three paused by David; round one's real state and Fotini's two campaigns read from the account's activity log; framework
clinical-evidence-traysaved (§4, §6, §8, §9). - 2026-10-30 (10:30 PT) — The winners read from Meta (AD025 · AD024 · AD034 · AD031 · AD033 — four of five = the news-chrome frame) and round three built as versions of them (§6, §8): run 2026-10-30-a, 28 cards 1:1, hooks batch 5, one re-render (r3-10, a real network mark), board /docs/nailscience/boards/statics-r3/, ask ns-statics-r3-judge; r3-concepts.json with the winners ledger.
- 2026-10-30 (09:42 PT) — Round three PAUSED by David in Ads Manager after day one's ~$22 (§6); the agent had switched rounds two and three on at 00:00:53; twin/doc/board/page carry it.
- 2026-10-29 (21:50 PT) — Weird statics r2 judged (7 keeps) and LAUNCHED as round three (§6, §8): campaign 120250102841410510, ABO $20, one ad set, AD051–AD057, the proven pair, ACTIVE at 00:00 Oct 30 with round two; twin v3.2 (round three, rounds renumbered, $170/day note) + doc + board/page re-rendered; copy sheet rebuilt; ask ns-statics-r2-judge answered.
- 2026-10-29 (17:25 PT) — Wild statics r1 judged (15 kills) and the 15 keeps LAUNCHED paused to NS-P2 (§6, §8): campaign 120250100703970510, CBO $50, two ad sets by story, AD036–AD050, the proven pair verbatim, ACTIVE at 00:00 Oct 30 with MGI + WAM; media plan twin v3.1 + doc (round two LAUNCHED, rounds renumbered) + board re-rendered; copy sheet rebuilt (live rows); the five surviving NEW formats saved to
creative/dashboard/image-frameworks.json; ask ns-statics-r1-judge answered (console update); round two of the weird statics on the saved frameworks (run 2026-10-29-b) rendered and boarded as statics r2 (ask ns-statics-r2-judge). - 2026-10-29 (17:00 PT) — Wild statics r1: 30 cards rendered, QA'd (four re-renders), on the hub board for keeps; media plan round two proposed (keeps → NS-P2 on the proven headline/body); ask ns-statics-r1-judge.
- 2026-10-29 (15:00 PT) — Truth Gate (53 claims) + Angle Studio r1 (12 angles · 94 hooks · bucket brief · copy sheet) built and boarded; asks ns-truth-gate + ns-angle-board-r1; the statics round waits on GO angles (§6).
- 2026-10-28 — Analytics live: https://launch-center-hq.netlify.app/nailscience/analytics/#pnl updates every hour (tools/analytics/README.md). The assumptions' EMPTY SKU prefix took the whole acm2 store, Charm's
CH-orders included →nailscience; "Try It First / Try it For Free" → the tbyb-sub lane. 69 orders since launch; Meta spend all kept (Liaison – A is NailScience's own account). - 2026-10-26 — attribution fixes (§6, §9):
ns-attribution.js(_ns_adq),build_checkout_tbyb.py(restore before the SDK + gate),shared/functions/order-webhook.mjs(3.55 full order, opt-in;partner_skuwhen on) +ENV_SETUP.md,set_prod_env.py(NEXT_FETCH_FULL_ORDER=1prod + staging),tools/qa/nailscience_tbyb_order.py(blocks tracking, renewal poll),verify_tracking.py(foreign case own number), new teststools/qa/order_webhook_fullorder_test.mjs+tools/qa/nailscience_adtouch_check.py; staging → production. - 2026-10-25 — the rows → Mild / Moderate / Severe Fungus (§3, §6):
offer_box.json,build_pdp_extras.py,build_pdp_read50.py,build_site_pages.py,build_checkout_tbyb.py,build_29next.py(→--align, packages renamed in place, map rewritten),shopify-spec.json(→--align); PDP chain + site pages + handoff stamp + compliance + checkout rebuilt; staging → production; live checks. - 2026-10-25 — the /dev checkout unblocked (David: "Do this task here: Unblock the /dev checkout pull"): /dev was 104 commits behind and
git pull --ff-onlyrefused on 17 files. The working tree held an uncommitted UNDO of the Oct 20 $24.99 change: 34 files byte-identical to Oct 14–18 commits (the PDP + checkout + 29next + Shopify builders and outputs, BRAND.md, RELAUNCH.md, the launch board, the sitemaps, the proforma + unit-economics data) while HEAD's last change to each was24360211a"the single kit is $24.99" — plus 7 hub renders my own Oct 20 build had made from that stale BRAND.md. All 34 restored to HEAD (lossless: every one was already in git history), /dev fast-forwarded to258ae50f0, the J&T dashboard back on 200. Kept untouched as real work: 7 CellScience prelander edits that exist nowhere else (avatar width/height on the comment threads),profit/mer/data/amazon-manual.json(today's MER automation), 4 files whose content lives only in a stash, and 160 tracked files someone deleted locally (97products/memo, 36tools/deep-researcher, 16tools/qa, 10adscaler). Live checked after: the PDP and checkout show $24.99, the hub brand DB has the Oct 20 rows — the stale copies never shipped. Backup of every tracked edit:~/Downloads/Code & Dev/_backups/dev-pull-2026-10-25/(the 34 files + a full patch). - 2026-10-24 — PDP announcement bar LAUNCH WEEK → BLACK FRIDAY (§6), source edit + staging → production.
- 2026-10-22 —
tools/checkout_selector.py: the handoff state is OPT-IN (a spec passes"handoff": {…}, mode defaults tohide) and its CSS / the no-ship rule are emitted only when used — another brand's rebuild stays byte-identical (CellScience's rebuild for its InitiateCheckout fix proved it); NailScience passes{mode: hide}and lost only five lines of unused one-row CSS. - 2026-10-22 — attribution values (§6):
build_checkout_tbyb.pyck_values()(SDK cart values, IC waits for the cart,NS_PKG_VALUEfallback, gate),build_pdp_extras.prices()(ViewContent + the cost-per-treatment cell = the single-kit price, gates); rebuild + PDP chain; staging → production; the attribution set saved intools/qa/results/nailscience-attribution-20261022/(probe + run log). - 2026-10-22 — checkout: package section hidden on
?pkg=, no struck retail, then no order summary (§6, §8):checkout_selector.pyhandoffmode(hide | collapse;sel3-section),build_checkout_tbyb.pyhandoff: {mode: hide}+strip_text()+ gates,build_order_summary.pySTRIKE = False+strip_text(),build_checkout_read50.pygate;nailscience_tbyb_order.pyreads the SDK cart; staging QA → production twice. - 2026-10-21 (afternoon) — offer box rows cleaned + trial row off + the clinicians chip (§6, §8, §9):
offer_box_v2.pyswitches (shipper tier,strike,trial_row) +offer_box.json;checkout_selector.pyflags (strike,ship_line,trial_pick) +build_checkout_tbyb.pySPEC; the chip markup + CSS in the PDP source[05b]; the PDP chain + the checkout rebuild; the same chip on CellScience v1 (its BRAND §6); staging QA → production. - 2026-10-21 — the checkout never re-asks the package + the Clinicians' Choice card off (§6, §8, §9):
tools/checkout_selector.pyhandoff state (sel3-pick·sel3-pick-sel· the "Your Package · Change" header pair · per-row pick CSS; the no-image class regex) →build_checkout_tbyb.pyrebuild + gates; the PDP source without [05b] + its CSS,build_pdp_read50.py/pagespeed.pywithout the card's rules, the PDP chain; the boardtools/brand-hub/docs/nailscience/clinicians-badge/board.html; askns-clinicians-badge; staging → QA matrix → production. Production proofs (2026-10-21 midday): the collapsed-selector check A–F ALL PASS on iPhone WebKit + desktop Chromium (screenshotstools/qa/results/nailscience-pick-20261021) · attribution in the browser ALL PASS · the webhook relay ALL PASS · ladder order 20/20 (ns-tbyb-20261021-103808) · trial monthly 26/26 (ns-tbyb-20261021-103839, the QA subscription cancelled) · the purchase-path matrix on the?co=shopifypreview 16/16 (nailscience-shopify-20261021-1039) · the prelander proof 41/41 (ns-prelanders-20261021-104247). Staging first, same set plus the single and trial one-kit lanes (20/20 · 26/26). - 2026-10-21 — the ladder on 29next again (§3, §6, §8): spec
handoff.on: false(+qa.query= the Shopify preview for the matrix),shop_handoff.pyre-stamp ×3,build_pdp_extras.pyre-seat;verify_tracking.pydormant branch;nailscience_tbyb_order.py --lane single; deploy staging +--prod; proofs live. - 2026-10-20 — the single kit $24.99 (§3, §6):
build_29next.py(UNIT_PRICE, LADDER, mapladder_pricesfrom LADDER, align sendsis_recurring) →--align;creative/nailscience/shopify-spec.json→shopify_store.py --align;build_pdp_extras.pyprices()+ gates, the OB-C plan line;build_checkout_tbyb.pyrows from the map;build_site_pages.py,pages_content.py,check_compliance.py,build_oto.py;tools/proforma/assumptions/nailscience.json+ the gate; the PDP chain ×2, site pages, checkout, OTOs rebuilt; deploy staging +--prod; checks live. - 2026-10-18 — the natives pass (§6):
ads/research/nail-fungus/natives/r1.json+ NATIVES-r1.md + OPERATORS-r1.md, boardnailscience--natives-r1, hub thumbsdocs/nailscience/natives/. - 2026-10-18 — no price on the Add To Cart button (§6):
build_pdp_extras.pyatc_labels(+ gates), the sticky / drawer titles on the April names; deploy staging +--prod; row-click check live. - 2026-10-17/18 — the April bundle names (§3, §6):
offer_box.json,build_pdp_read50.py,build_site_pages.py,build_checkout_tbyb.py,build_29next.py(LADDER names + RENAME_FROM accepting earlier names) → packages 2/10/11 renamed in place,creative/nailscience/shopify-spec.jsontier names →shopify_store.py --align,tools/qa/nailscience_tbyb_order.py(finds the ladder package by id); PDP chain + site pages + checkout rebuilt; deploy staging +--prod. - 2026-10-16 (night, 10:41 PM PT) — round one LIVE: the four April winners switched on untouched (§6); the plan twin + doc dated and on $100 a day;
ns-ads-goanswered,ns-third-slotposted. - 2026-10-16 (night) — the go given in chat; preflight passed (§6); the switch-on refused by Meta (asu has no Advertise permission on Liaison – A, §4); nothing switched.
- 2026-10-16 — Media plan v3: R1 slot C = three videos (V12 + two), R2/R3 fresh slots = seven of the 54 keeps; the 30% page gate on the board.
- 2026-10-16 — the April prelanders at their April URLs + one count per action (§4, §6, §8, §9):
tools/deploy-health-report.sh(stages NS-P3 at / and NS-P2 at /blog/ranking/, relay + pixel + PDP-link gates, smoke on the winners' exact links; draft → production ×2);prelanders/drfranks+prelanders/nailscience/top5-advertorial(relay on that host, buttons → PDP, sticky refresh, Pattern M) + the brand-site mirrors;tools/qa/nailscience_prelander_proof.py(new, 41/41 live);shared/shopify/shop-handoff.jsv2 (store_reports,bound,report) +tools/shop_handoff.py+ the spec'sstore_reports: ["meta"];build_pdp_extras.pyatc_wiring(Ladder A AddToCart values);verify_tracking.py(GA4 capture through the handoff, the store-reports expectation, the[shopify]checkout check);deploy.shstaging +--prod; media plan md/json (the destination decided, decisions 1–2 closed), pages manifest (NS-P2/NS-P3 live_url = the April URLs, mirror_url = the copies). - 2026-10-16 — The media plan board switches launches (David: "Let me switch through the different launches by clicking on that"): the brand name lists every launch with a board. This board also wears NailScience's own color now (it showed CellScience red). Account read Oct 16, 8:57 AM: nothing active, $0 today and yesterday; W1–W4 still paused, waiting on the go.
- 2026-10-15 (night) — the trial lane takes orders (§3, §6, §8, §9): the $6.99 store method created in David's Chrome (29next → Settings → Fulfillment),
build_29next.py --createattached it as ref 2,build_checkout_tbyb.pyre-stamped (trial card on ref 2),deploy.shstaging +--prod; dummy-card trial orders monthly 26/26 + one-kit 26/26 (staging) and monthly 24/24 on production iPhone (order 104892; subscriptions 929 / 930 read $0 shipping, then cancelled);tools/qa/tbyb_shipping_sweep.pygains--brand nailscience(dry run: 0 trial-born subscriptions carrying shipping); the runner prints the VIP renewal state; David's chat answers on the hub; asks withdrawn / re-posted. - 2026-10-15 (night) — Media plan board at
/docs/nailscience/media-plan/board.html(David picked the Schedule design, CellScience BRAND §6): the three rounds counted from the go, W1 + W2 in the Foot Doctor campaign slot and W3 + W4 in the Top 5 slot (shared CBO slots), V12 in its own; the April numbers shown as the April run, never as this plan's spend; the three asks as what waits on you. Twin: titles, April verdicts,rounds[].slots, page names,board(the $4 checkout bar, the rules in plain words, the decisions). Refresh:python3 tools/brand-hub/media_plan_board.py all --brand nailscience. - 2026-10-15 (midday) — PRODUCTION + the QA stack (§6, §8):
deploy.sh --prod×2 (11:20 the relaunch build; 12:00 the paint-first boot);build_oto.py(the VIP rebuild + the shim on upgrade / bundle / thank-you; chassis assets inspecial-offer/assets/; the 78 scraped vendor files retired);build_checkout_read50.pyfolded intobuild_checkout_tbyb.py(+ the CSS bundle, Swiper removal,<picture>hero,defer, paint-first);tracking_head.paint_firston every page; the runtime's Patterns D + M;shared/shopify/shop-handoff.jsInitiateCheckout (vendored);verify_tracking.py(browser + webhook proofs; signs with the webhook's own key);tools/qa/nailscience_tbyb_order.py(lanes sub / one / ladder; typed card; every API write captured);tools/qa/cellscience_tbyb_order.py(typed card);set_prod_env.py --webhook-secret-from-store(+--mirror-webhook-to-staging);tools/sync-blog-mirrors.sh/blog/exclude; the pages manifest's live URLs + dates; results intools/qa/results/(nailscience-shopify-20261015-1116 / -1121, ns-tbyb-20261015-1133 / -1143 / -1201 / -1206). Memory29next_checkout_qa_traps.md. - 2026-09-02 — created from template + imported 3 memory files as the starting record.
- 2026-10-14 (evening) — Media plan 2026-10 written (§6): the twin + the plain-language plan + the hub page (tabs per round, the pair that bought marked on every winner, V12 playable),
tools/ad_codes.py on(switch existing ads on, siblings paused first — dry run shows 7 pauses + 2 campaign switches), the ad-codes register backfilled, NS-P3 coded in the pages manifest, the two April prelanders mirrored intoproducts/nailscience/blog/,tools/deploy-nailscience.shwritten (draft deploy proven: hygiene PASS, both mirrors + legal + checkout 200, relay 400 — preview6ad01e5236de7061b5dfa9c3--vocal-sunburst-5f0620.netlify.app; production untouched), the old domain's guarded 301 lines intools/deploy-health-report.sh, the Top 5 page 60-day → 30-day, V12 moved tocreative/nailscience/videos/and hosted on the UGC Studio, the launch board file, three asks on the lane. - 2026-10-14 (night) — the 50+ readability pass on every page (§6):
build_pdp_read50.py, the chrome/collection/legal type and accent fixes, staging redeployed; PDP measured 0 findings at 375/390, home/collection/legal 0. - 2026-10-14 (late) — Hero carousel pass (§6):
build_pdp_extras.py(zoom · looping arrows · WebP twins · slot ids · the sticky urgency line), 30 twins in/assets/,build_pdp_read50.pysizes the arrows; the [13] urgency block removed; staging redeployed, QA'd at 390/1280. - 2026-10-14 (late) — Try Before You Buy lane (§3 table, §6):
build_29next.py(products 202–208, packages 10–13, pkg 2 renamed, the map),build_checkout_tbyb.py+build_order_summary.py(the checkout stamp),build_pdp_extras.py(the?ob=ccolumn + the package wiring),set_prod_env.py --applyon production (29next admin base/key, the SKU prefix gate, the lane names),netlify.toml+_redirects(the/next-api/*proxy); asks:ns-tbyb-shippingposted,ns-campaign-reactivatewithdrawn; staging redeployed. - 2026-10-15 — Shopify ladder live on staging (§3 Shopify block, §6):
creative/nailscience/shopify-spec.jsonfinalised,shopify_store.py --create / --shipping / --storefront-token, the handoff on the PDP (throughbuild_pdp_extras.py, which lifts and re-seats the handoff tags so re-runs stay byte-identical and runs the handoff's--check), the home and the collection;shop_handoff_check.py24/24,shopify_matrix.py --quick16/16 on staging; asksns-shopify-finishposted /ns-shopify-storewithdrawn; staging redeployed. - 2026-10-14 (evening) — Phase 4 on staging (David: "get the offerbox looking right with the shopify + 29next … readability and sitemap and homepage stuff and footer"): the PDP moved to
/products/nail-restoration-kit/with the radio offer box on Ladder A + the trial row, the guarantee card, the site footer, absolute asset paths, canonical + OG + Product/Breadcrumb JSON-LD, the word "mechanism" gone, the long sentences split (COPYCHIEF-pdp-r1.md); homepage + collection + seven legal pages + robots + sitemap built from the ported CellScience builders;netlify.tomlrewritten (pretty legal URLs 301 →/pages/<slug>/, old.htmlpaths 301,/?pkg=+/?ob=301 → the PDP, funnel pages noindex, asset caching) +_redirects/_headersregenerated;deploy.sh(gated, staging + prod); the staging site created; the May legal.htmlfiles andvercel.jsonretired; the checkout/OTO guarantee swept to 30 days + the 30d badge. Found: the 29next campaign is inactive (§8). Asks posted:ns-campaign-reactivate,ns-staging-review,ns-shipping-promise. - 2026-10-14 (later still) — Deep-research chain cancelled on David's call (§6); the generated prompts, chain script and logs removed from
tools/deep-researcher/runs/nailscience/at-ns/, the April brief restored as07_copy-brief/response.md. Research base = that brief (Mode A). - 2026-10-14 (later) — Ladder A locked (§3): assumptions file moved to A and the gate re-run (FIX at $60, passes under $43.16 — the CPA target); scratch candidates deleted;
creative/nailscience/shopify-spec.jsondrafted from the ladder for Phase 4 (waits on the store, askns-shopify-store); lane state + hub redeployed. - 2026-10-14 — Seven verdicts pulled from the Launch Center (§6) and filed into §0–§4; asks
ns-shopify-store(the owner's five Shopify steps) andns-offer-pick(ladders A/B/C through the gate: FIX / FIX / PASS at $60) posted; the April copy brief parked as07_copy-brief/response.v0-april.mdand the research chain 01→07 queued on the Gemini driver (tools/deep-researcher/runs/nailscience/at-ns/run_chain.sh). - 2026-10-13 — RELAUNCH kickoff. Index + plan filed (
RELAUNCH.md); §0–§9 populated from the repo, the live sites, Netlify, the 29next store and the May profit data; brand re-registered intools/brand-hub/brands.json(in-launch); console lane opened with 8 asks; economics gate run (KILL at $60 · passes under $35.30); the template→PDP overlay row removed fromtools/sync-blog-mirrors.sh. May 2026 run on record: spend $1,637.35 · 26 orders · $1,362.23 gross ($1,143.40 front end + $218.83 upsells) · CPA $62.98 · AOV $52.39 · CTR 11.0% · CPM $47.21 · campaigns "J&T – Foot Doctor Reveals" + "J&T – Top 5 Rankings" · 14 × 6-pen, 10 × 9-pen, 2 × 3-pen.
Imported status records (2026-09-02) — history. The facts were distributed into §0–§10 on 2026-10-13; these records stay as the April–June record.
From memory nailscience_klaviyo_bootstrap_brief.md
NailScience — Klaviyo Bootstrap (Session Brief)
This is the canonical handoff. Read this first. NS Klaviyo was diagnosed virgin/pre-bootstrap 2026-05-19 — there's no built infrastructure to "triage" yet; this session builds it with every HF + WAM lesson already in the playbook so we don't re-discover them.
Scope
IN
- Bootstrap NS Klaviyo end-to-end (HF-pattern): commerce integration → backfill profiles + events → segments (native-metric) → popup → welcome flow → recovery cascade (cart / leads / lifecycle).
- Wire
tools/29next-to-klaviyo/backfill.pyfor--brand nailscience(config slot at line ~103 is a placeholder today; fill it). - Bootstrap
tools/29next-to-klaviyo/create_segments.pyfor NS (NS prefix already registered in BRAND_PREFIX dict). - Bake the proven HF/WAM patterns from day one — do NOT rebuild any of the bugs we already fixed.
OUT
- NS attribution / tracking work (RESOLVED 2026-05-16 + Phase-3.5-VERIFIED — see
nailscience_tracking_status.md). Don't touch. - HF and WAM Klaviyo work (different sessions, different brands, separate creds).
- Anything that requires the WAM cascade-preview Netlify site (siteId
6dfa0488-…) — that's a WAM site.
Current NS Klaviyo state (verified 2026-05-19 — empty)
- Org: Nail Science ·
RA2aJd(private key in~/.klaviyo/nailscience.env, mode 600). - Lists (3):
Y4EDQuEmail List ·SAKJXDPreview List ·V6PVs9Text Messaging List. Real members ≈ 0–1 (clean start). - Segments (4) — ALL Klaviyo defaults: Engaged 30/60/90 Days + New Subscribers. Zero custom segments.
- Metrics (20) — ALL Klaviyo-native engagement +
Active on Site. Specifically MISSING:Placed Order,Started Checkout,Fulfilled Order,Viewed Product,Added to Cart— i.e. no commerce integration is sending order events into Klaviyo. This is the #1 thing to fix. - Flows: 0 (confirmed across 5 list-query shapes).
- Templates: none (since no flows).
The diagnosis is not "bugs to fix" — it's "infrastructure to build."
Adjacent context (read for orientation, don't touch)
- NS 29next merchant subdomain:
acm2.29next.store(pernailscience_tracking_status.md: "storeacm2, ID 67, secret1ef14a260093"). The 29next webhook there is part of attribution; if you wire Klaviyo's Shopify-style integration via 29next admin webhooks/API, expect to coordinate around the existing attribution webhook. - NS domain:
trynailscience.com(PDP),quiz.trynailscience.com(quiz),thehealthreportdaily.com/drfranks(advertorial). - NS Meta pixel:
1341229411229569(≠ HF pixel — never cross-fire). - NS PDP canonical:
template/pdp/02_ecom/index.html(overlay copies intoproducts/nailscience/index.html; edit canonical, sync writes overlay). - NS niche: nail-fungus. Subject-line manifest scaffold exists (empty) at
email/nailscience/subject-lines/manifest.json.
Bootstrap playbook (the actual work)
Phase 0 — Decide commerce integration mode
- HF used 29next backend (
spinerelief.29next.store) withbackfill.pydoing a bulk profile + Placed Order + Started Checkout import via the 29next admin API. - NS uses 29next too (
acm2.29next.store). Same playbook applies. Use the existingtools/29next-to-klaviyo/backfill.pyscript. - Alt: if NS has a Shopify storefront in addition to 29next, Klaviyo's native Shopify integration is the simpler path (auto-creates Placed Order / Started Checkout / Fulfilled Order metrics). Check with David before committing.
Phase 1 — Wire backfill for NS
Extend tools/29next-to-klaviyo/backfill.py BRANDS dict (currently line ~99–106), filling the commented placeholder:
```python
"nailscience": {
"next_admin_base": "https://acm2.29next.store/api/admin",
"klaviyo_env_file": Path.home() / ".klaviyo" / "nailscience.env",
"next_env_file": Path.home() / ".29next" / "nailscience.env", # needs to be created
"default_list_id": "Y4EDQu", # Email List
"checkout_url_host": "trynailscience.com", # or the actual 29next checkout host
"internal_email_domains": ["trynailscience.com", "ecomscalerslimited.com"], # confirm w/ David
},
```
- NS 29next admin token: needs to be created/dropped in
~/.29next/nailscience.envasKLAVIYO_PRIVATE_KEY(29next bearer token; ask David — same shape as HF'sYAXhL8lf...). - Run
backfill.py --brand nailscience --dry-runfirst; review output; then--apply. - HF lessons baked in: PATCH-on-409 for existing profiles, role-account/domain exclusion, guest-checkout scaffolding (already in backfill.py).
Phase 2 — Build segments with NATIVE METRICS (NOT properties)
The single most important lesson from the HF triage: the original HF bootstrap gated segments on properties['29next_orders_count'] / 29next_user_type / 29next_total_spent — these are written ONLY by the ongoing sync and FAIL SILENTLY (events keep flowing, segment quietly stops growing). The Klaviyo metric Placed Order is native (fired by the integration) and works regardless.
Build NS segments natively from day one — don't repeat the property-gated mistake:
- NS Customers = Placed Order count > 0 (in-the-last 1825d)
- NS Leads = Placed Order count == 0 AND Group2 = OR(
properties['ns_signup_source']existence:not-set,properties['ns_signup_source']not-equals 'pdp_exit_popup') — the popup-arc carve-out from day one, so the deep-discount Leads flow never blasts a fresh popup capture. - NS VIP = Placed Order count > 3 AND sum > $299.99 (Klaviyo
measurement: "sum"sums the metric's$value). - NS Active Cart Abandoners = Started Checkout > 0 AND Placed Order == 0 (last 30d).
tools/29next-to-klaviyo/create_segments.py currently uses the OLD property-based pattern — update it before running for NS (or write a NS-specific bootstrap script using the native-metric pattern). See klaviyo_segment_api_skill.md for the corrected pattern + authoritative segment schema.
Phase 3 — Build popup + welcome flow (HF pattern)
Mirror products/humanframe/shoulder/index.html markers [33b]/[35]/[37] in the NS PDP. Property stamp properties: { ns_signup_source: 'pdp_exit_popup' }; segment NS Popup Captures matches on equals. Welcome flow triggers on that segment, 5-min delay, plain-text-personal E1 (Maria-equivalent named sender — pick the NS persona).
Producer/consumer check (HF lesson): the popup property KEY and the segment property KEY must match EXACTLY. Verify before flipping NS_EMAIL_POPUP_ENABLED=true.
Phase 4 — Build the cascade / recovery flows
HF Cart (4-day akemi, 83% RECOVER83) + Leads (4-day akemi, 88% RECOVER88 with full-akemi education body) — same shape. WAM cascade (#02–#05 4-flow funnel-stage cascade, flat 20% WELCOME20) — alternative shape. Pick the one that fits NS's offer ladder.
Apply ALL these hard rules from day one:
HARD RULES — apply throughout (do not re-discover)
1. Klaviyo Liquid URL-encode filter = urlencode, never url_encode. One bad filter fails the WHOLE template render silently — the HTML looks perfect, only POST /api/template-render/ surfaces the error.
2. Verify every email template via POST /api/template-render/ — reproduces Klaviyo UI Preview. Visual / GET inspection does NOT catch render-breaking Liquid. Same class as "verify segments by membership size."
3. Verify segment edits by MEMBERSHIP SIZE, not HTTP 200 alone. A schema-valid def can be semantically empty (HTTP 200) — GET /segments/{id}/profiles/?page[size]=100 after a ~60–90s recompute wait and assert sane count.
4. Klaviyo not-equals excludes profiles where the prop is UNSET. To "exclude only a tagged cohort, keep the unset cohort," OR an existence filter: {conditions: [{type:profile-property, ..., filter:{type:"existence",operator:"not-set"}}, {type:profile-property, ..., filter:{type:"string",operator:"not-equals",value:"X"}}]}.
5. Segment defs are PATCH-able in place (PATCH /segments/{id} → 200, keeps id, no flow re-trigger needed). Flow content is NOT — PATCH /flows/{id} 400s on definition; only status+name mutable → flow content changes = delete+recreate.
6. Segment defs CANNOT reference other segments. profile-group-membership.group_ids accepts LIST ids only ("Group does not exist" error on segment ids).
7. profile-filter timeframe: flow-start ≠ alltime. flow-start = exit-on-new-conversion (count from flow entry); alltime = exclude-prior-converters (count over all history). Discount/scarcity recovery flows almost always want alltime — flow-start lets past customers RE-enter the deep-discount flow if they start a new checkout.
8. Flow create schema: actions need temporary_id + links:{next} chaining + top-level entry_action_id. A time-delay with unit ∈ minutes/hours must NOT include delay_until_weekdays/delay_until_time (GET echoes them but create 400s — strip when unit != 'days').
9. Recreate verification = render-gate + structural parity. Before deleting the old flow, assert the new flow's triggers, profile_filter, action-type sequence, delay values, and subjects all == the saved old def — render-OK alone doesn't prove faithful chaining/timing.
10. Honest-discount HARD RULE: any code referenced in any send must be REAL in 29next/Shopify with a real expiry. False codes destroy the list. Placeholders OK in DRAFT; replace with real codes before activation.
11. /usr/bin/curl -g (or --globoff) for any Klaviyo URL with [ ] (fields[…], additional-fields[…], page[size]). Python stock urllib has no CA bundle → shell to /usr/bin/curl.
12. Read the OpenAPI (https://raw.githubusercontent.com/klaviyo/openapi/main/openapi/stable.json) for any unfamiliar schema — components.schemas.ConditionGroup for segment conditions, FlowCreateQuery+FlowDefinition+SendEmailAction+TimeDelayAction+Link for flow create, AlltimeDateFilter/SinceFlowStartDateFilter/ExistenceOperatorExistenceFilter for the timeframe/existence shapes. Brute-probing wastes hours; the API's "invalid field type" 400s never list valid values.
13. Klaviyo metrics list endpoint quirk: /metrics/?page[size]=100 may silently return 0; use /metrics/ (no params) or direct GET by id. Same for /flows/ — list may return 0 even when flows exist; if you know specific ids, GET directly.
14. MutationObserver HARD RULES apply to any NS PDP popup code — never write to the DOM inside a callback without a guard flag; split control listeners (close/X/Esc/overlay) from trigger listeners (exit-intent/dwell) so teardown() doesn't nuke close handlers. See CLAUDE.md mandatory checks + the HF popup close-bug fix d77a443.
15. Klaviyo Client Subscribe API returns 202 even if target list is deleted — failures are silent. Before deleting any list, grep-audit ALL references (popups, flow triggers, segment defs, scaffold scripts).
16. Offer-ladder collision check is REQUIRED whenever you change any segment definition that's a flow trigger. Re-run the suppression analysis; for single-consumer trigger segments, suppression goes ON the SEGMENT (in-place PATCH, no flow rebuild) + honest rename. See klaviyo_segment_api_skill.md "Offer-Ladder Lesson."
17. "Powered by Klaviyo" branding is plan-dependent — Klaviyo auto-appends it to delivered emails on free/non-paid plans. Removable in Account → Settings → Branding on paid plans. Verify before flipping any flow live; a "Powered by Klaviyo" badge breaks plain-text-personal voice instantly.
18. Double-opt-in on the capture list holds the profile pending until confirmation — the property may not persist + segment may not fire until confirmed. Verify the NS Email List Y4EDQu opt-in setting before flipping the popup live.
Credentials + environment
- NS Klaviyo private key:
~/.klaviyo/nailscience.env(set up 2026-05-19; mode 600).KLAVIYO_PRIVATE_KEY=pk_RA2aJd_…,KLAVIYO_PUBLIC_KEY=RA2aJd. REVOKE after rollout (add to the standing revoke list inhumanframe_klaviyo_status.mdaction items). - NS Klaviyo MCP connector — may or may not be set up. The
mcp__8cd8eef9-…__klaviyo_*connector earlier in the session was the HF org. For NS, you'll likely need REST directly (which works fine). - NS 29next admin token: NOT YET PRESENT. Will need
~/.29next/nailscience.envwithKLAVIYO_PRIVATE_KEY=<bearer>(29next bearer token, formatYAXhL8lf…-style). Ask David. Fromnailscience_tracking_status.md: storeacm2, webhook ID 67, secret1ef14a260093— these are attribution webhook creds, NOT necessarily the admin token; David has the admin token separately. - Repo locations: - PDP canonical:
template/pdp/02_ecom/index.html- PDP overlay:products/nailscience/index.html(synced from canonical — edit canonical not overlay) - Email dir:email/nailscience/(currently onlysubject-lines/manifest.json) - Backfill tool:tools/29next-to-klaviyo/backfill.py(BRANDS dict line ~99–106 — extend for nailscience) - Segment tool:tools/29next-to-klaviyo/create_segments.py(BRAND_PREFIX has'nailscience': 'NS'— but the segment defs use the OLD property pattern; rewrite to use native metrics per Hard Rule #3 above)
Cross-references
humanframe_klaviyo_status.md— the proven HF bootstrap pattern + every trap caught + every fix applied. Read this; NS bootstrap is the same shape minus the bugs.klaviyo_segment_api_skill.md— authoritative segment condition catalog + filter operator enums + all HARD RULES on segment schema (PATCH-able, not-equals/unset trap, etc.).email_copywriting_skill.md+ reposkills/email-copywriting/SKILL.md— voice + cascade-by-intent + akemi escalator + 3-day promo cadence + the deliverability/format rules.nailscience_tracking_status.md— NS attribution (RESOLVED, Phase-3.5-verified). Read for brand context + the 29next store reference (acm2, webhook ID 67) + the Pattern I attribution pattern that's already deployed to all NS paid paths.wam_klaviyo_status.md— alternative bootstrap pattern (Shopify-native integration vs 29next). May or may not apply for NS.wam_campaigns_session_brief.md— example of the sibling handoff pattern for a different brand's email session.- Obsidian:
Skills/Klaviyo Flow Build & Checkout Prefill.md— flow-build patterns + traps + the popup chain pattern. CLAUDE.mdmandatory checks — MutationObserver audit rules; HARD RULE on Shopify Liquid{% comment %}vs HTML<!-- -->if you touch any Shopify theme.
First actions for this session
1. Read this brief (you just did).
2. Read humanframe_klaviyo_status.md — the playbook you're about to apply, plus what NOT to repeat (the property-gated segment trap, the flow-start filter trap, the url_encode render-bug trap, the popup close-bug trap, etc.).
3. Read klaviyo_segment_api_skill.md — authoritative segment schema; apply native-metric pattern from segment #1.
4. Confirm with David:
- NS commerce backend: 29next (acm2.29next.store) or also Shopify?
- 29next admin token availability (need ~/.29next/nailscience.env).
- NS brand voice / sender persona for emails (HF used "Maria, HumanFrame Support"; WAM used "WAM!™ Mints / support@wammints.com" — what's NS?).
- Offer ladder: what discount codes are real in 29next/Shopify today? (Honest-discount HARD RULE — no flow ships with placeholder codes.)
5. Then start Phase 1: extend backfill.py BRANDS dict for NS, dry-run, review.
Hard discipline (apply throughout)
- Pre-compaction: before any context compaction, update Skills + Notes — system-wide rule. New NS lessons → into this brief or a new
nailscience_klaviyo_status.md(write the latter once there's something live). - Rebase → CMP discipline: rebase onto
origin/mainBEFORE any commit AND before any push. Surgicalgit addonly (never-A;creative/dashboard/index.htmlis hook-dirty and stays out). Force-with-lease for feature branches only, never main. - Deploy target verification: if you deploy anything to Netlify, verify siteId via
GET https://api.netlify.com/api/v1/sites/{siteId}before AND after. The wrong-site trap burns whole sessions. - No HF/WAM contamination: different envs, different orgs, different lists — never run an HF/WAM script against the NS account or vice versa. Always
source ~/.klaviyo/nailscience.envexplicitly.
From memory nailscience_klaviyo_status.md
Status (last updated 2026-05-20): Bootstrap COMPLETE + copy final — 3 flows DRAFT (CURRENT ids Welcome Yj86A3 / Cart U2C2B4 / Leads WTPuDZ). Codes WELCOME20/RECOVER83/RECOVER88 w/ 13-cap; bodies at FK ~7; Dr. Franks voice; dashboard wired. NS work UNCOMMITTED (David: skip git). Awaiting David's 29next code creation (both RECOVER capped at 13) + Klaviyo settings before flip-live.
NS Klaviyo went from virgin → fully bootstrapped in one session. Backfilled 53 profiles + Placed Order/Started Checkout metrics from 29next; built 4 native-metric segments; drafted + shipped 3 flows (Welcome / Cart / Leads) all DRAFT with render-verified templates. Only thing between current state and LIVE = David creating the 3 discount codes in 29next per honest-discount HARD RULE.
Klaviyo account
- Org: Nail Science · public key
RA2aJd· private key~/.klaviyo/nailscience.env(mode 600). REVOKE after rollout. - 29next merchant:
acm2.29next.store· admin bearer at~/.29next/nailscience.env(mode 600). REVOKE after rollout. - Brand domain:
trynailscience.com(checkout proxied on apex:/c/,/checkout/). - NS Meta pixel:
1341229411229569(attribution — separate concern, RESOLVED pernailscience_tracking_status.md).
Lists
| id | name | members |
Y4EDQu | Email List (popup/welcome target) | 54 |
SAKJXD | Preview List | 1 |
V6PVs9 | Text Messaging List | 0 |
Native commerce metrics (created 2026-05-19 by backfill)
| id | name | integration |
S7mTbV | Placed Order | API (from backfill, ongoing via re-runs) |
UPiQy6 | Started Checkout | API |
| — | Fulfilled Order | NOT created yet (no order-fulfillment events being pushed; non-blocking for Cart/Leads; needed only if you build a Review/UGC post-purchase flow like WAM #08) |
Plus 20 Klaviyo-native engagement metrics (Opened/Clicked/Bounced Email + SMS variants + Active on Site).
Segments (4) — all NATIVE-METRIC (HF property-gated bug not repeated)
| id | name | def | members |
Y6Q6gc | NS Customers (have ordered) | Placed Order S7mTbV count > 0 (1825d) | 30 |
RZU8LJ | NS Leads — recovery (no order, excl. popup arc) | count == 0 AND OR(ns_signup_source not-set, not-equals 'pdp_exit_popup') | 22 |
Vw8Dsi | NS VIP Candidates | count > 3 AND sum > $299.99 (1825d) | 0 (legitimately — early store) |
Vrvp6U | NS Active Cart Abandoners | Started Checkout > 0 AND Placed Order == 0 (30d) | 14 |
Plus Klaviyo defaults (Engaged 30/60/90, New Subscribers) — left alone.
Popup-arc carve-out baked in from day 1 even though no popup is built yet — when one is built and stamps ns_signup_source=pdp_exit_popup, the Leads flow correctly excludes those captures without any segment edit. Saves the HF mistake we'd have re-made.
Flows (3, all DRAFT) — ⚠️ CURRENT IDs (rebuilt 4× this session — see history)
Flows were delete+recreated 4 times as copy evolved (each recreate = new flow id + new clone ids; standalone seeds were PATCHed in place and REUSED — they never changed id). Use ONLY the CURRENT row.
| CURRENT id | trigger | profile_filter | cadence | clones (current) | |
| Welcome | Yj86A3 | list Y4EDQu (Added to Email List) | none | delay 5m → E1 | VTFasV |
| Cart Abandon (4-day, akemi, 13-cap) | U2C2B4 | metric UPiQy6 Started Checkout | Placed Order S7mTbV count == 0 (flow-start — NS Cart still recovers returning customers re-abandoning; the alltime exclusion is only for HF-style once-only deep launch) | delay 2m → D0 → 1d → D1 → 1d → D2 → 1d → D3 | TpAXYr/SLzVmN/Y9RzTf/XwQxgC |
| Leads (4-day, akemi, 13-cap) | WTPuDZ | segment RZU8LJ | Placed Order count == 0 AND Started Checkout count == 0 (flow-start, both as exit) | D0 immediate → 1d → D1 → 1d → D2 → 1d → D3 | SFWF2q/WgNs3n/S4VCpM/VDMpqK |
Standalone seed templates (urlencode-clean; PATCHed in place across all rebuilds, ids stable): Welcome SksDTc; Cart VL3Udn/RtxQep/UkKihk/XQXNWU; Leads X9jvYT/X4GfDv/WY4Zi7/V6ZRse.
Rebuild history (superseded ids, all DELETED — do not reference): (1) initial build Welcome WNNuBj / Cart YwB7Lq / Leads XNWNNp (codes WELCOME15/RECOVER25/RECOVER30); (2) code swap to HF-naming Welcome WaQrSU / Cart Svmqwy / Leads XaYL5Y (WELCOME20/RECOVER83/RECOVER88); (3) 13-cap framing Cart RqHPJX / Leads WCJXzC (Welcome unchanged WaQrSU); (4) 7th-grade tighten → CURRENT Yj86A3/U2C2B4/WTPuDZ.
Copy state (CURRENT): codes WELCOME20 (20%, 14d, NO cap — soft welcome) / RECOVER83 (83%, 13-cap) / RECOVER88 (88%, 13-cap, separate pool) — HF naming per David. HF-parallel "13 launch codes" scarcity framing across all 8 Cart+Leads bodies ("we just launched the list-only program, capped at 13"). All 9 bodies tightened to FK grade 2.2–4.5 (7th-grade target, David). Dr. Franks voice throughout.
Sender: Dr. Franks / support@trynailscience.com (reply-to same). Smart Sending on. All 9 clone templates render-gated via POST /api/template-render/ — passed before flow creation; re-verified after.
Voice + brand
- Persona: Dr. Franks, podiatrist who wrote the
thehealthreportdaily.com/drfranksadvertorial. First-person, compassionate-authority register. Mechanism-anchored (2-step kit: Step 1 softens nail top layer, Step 2 antifungal goes underneath — the whole differentiator vs Kerasal/Kerassentials/$400 prescriptions). - Audience: women 50+, embarrassed about thick/yellow toenails. 40,000+ customers per prelander social proof, 3,214 Trustpilot reviews.
- Realism: flows reference the slow real-life nail-growth timeline (weeks 1-3 invisible, 4-6 new growth clear, 8-12 line obvious, month 4 cut away last infected) — counters "quit at week 6" failure pattern of most patients.
- Honest medical integrity: Welcome E1 P.S. refuses sale to anyone with serious symptoms ("completely black or detached at the bed, don't order"); Cart D3 P.S. offers to identify non-fungus issues. Differentiates from typical DTC trust posture.
⚠️ ACTIVATION GATES (David's, before flipping draft→live)
1. Create 3 real discount codes in 29next (acm2) — HF naming (David):
- WELCOME20 — 20% off first order, 14-day expiry, NO usage cap (soft welcome, no scarcity claim)
- RECOVER83 — 83% off cart recovery, Usage Limit = 13 (13-cap copy is in the bodies), ~5-day expiry
- RECOVER88 — 88% off leads recovery, Usage Limit = 13 (separate pool from RECOVER83), 7-day expiry
Honest-discount HARD RULE + 13-cap extension: the bodies say "13 launch codes, first come first served." If the codes are NOT capped to 13 redemptions in 29next, the scarcity claim is false the moment a 14th person redeems. Cap BOTH RECOVER codes to 13. Body copy refers to specific %; codes must match.
2. Confirm 60-day money-back guarantee is the live policy (drfranks advertorial says 30d; trynailscience.com prelanders say 60d — Cart D2/D3 + Leads D2/D3 reference 60). If wrong, swap in bodies.
3. Klaviyo account settings (Account → Settings):
- Account default sender / from-name (don't ship with default placeholder)
- Klaviyo branding ("Powered by Klaviyo" footer) — plan-dependent; if on free, it appears in delivered emails and undermines Dr. Franks plain-text-personal voice. Remove on paid plan.
- Account address / footer postal address (CAN-SPAM requires a physical address in commercial emails).
- Email List Y4EDQu double-opt-in setting — if ON, Welcome trigger may silently never fire (HF lesson). Verify OFF or that property persists pre-confirmation.
4. Domain authentication for trynailscience.com (DKIM/SPF/DMARC via Klaviyo Settings → Domains and Hosting). Required for Primary-tab deliverability.
After all 4 gates clear: PATCH /api/flows/{Yj86A3|U2C2B4|WTPuDZ}/ with attributes.status: live and they start sending.
Dashboard (Email J&T) — NS wired 2026-05-19
- Served by
tools/jt-server.py --port 8770→http://127.0.0.1:8770/email/dashboard/. NS appears in the brand selector (from/email-runs/__list). - NS manifests:
email/nailscience/bodies/manifest.json(9 bodies, per-emailarc_role+ 13-cap context) +email/nailscience/subject-lines/manifest.json(9 subjects w/ framework tags). - Flow Library canonical-ID fix (David caught "linking wrong flows to wrong emails"): the dashboard's
FLOW_LIBRARYinemail/dashboard/app.jsis the canonical Klaviyo taxonomy (01-welcome,04-abandoned-cart,05-abandoned-checkout, …). Manifestflow_idMUST match a canonical id or the Flow Library mis-groups. NS Cart trigger is Started Checkout → canonical05-abandoned-checkout(not "cart"); NS Leads →12-leads-recovery(NEW canonical entry I ADDED toFLOW_LIBRARYfor the list-only recovery pattern HF+NS share). Remapped both NS manifests. HF's manifest uses non-canonical ids (cart-abandon/leads/welcome) — same latent bug, David said LEAVE ALONE. Dashboard cache-bust bumped to?v=26inemail/dashboard/index.html.
⚠️ Git state (2026-05-20) — NS work UNCOMMITTED by David's choice
Rebase/CMPD was attempted but origin/main is 1013 commits ahead with 2 months of newer WAM development. My 17 May-era commits (HF+WAM email) hit deep conflicts on ~13 WAM cascade bodies (main did ™→® migration + honey/lemon/ginger-flavor-not-ingredient rule + relative-SVG vs my Klaviyo-CDN-PNG image hosting). WAM cascade is LIVE → not force-resolved. David chose "skip git entirely — just Obsidian + Skills." So: NS bootstrap is on-disk UNCOMMITTED (backfill.py NS BRANDS entry + 9 NS bodies + 2 NS manifests + dashboard app.js/index.html). Rebase aborted cleanly, stash popped, tree restored to pre-attempt state. My unique WAM 08/09 bodies are NOT on origin/main (would be lost in a take-theirs rebase) — flagged for the eventual reconciliation pass. NS files are all-new (zero main conflict) so committing them later is low-risk; the WAM/main divergence is the real reconciliation project, deferred.
What's NOT built (deferred / out of session scope)
- No PDP popup. David's call (Phase-0 answer: "skip popup; bootstrap email/flows only"). When/if added later, the segment carve-out is already in place — popup just needs to stamp
properties.ns_signup_source='pdp_exit_popup'. - No post-purchase / Review / Winback flows. WAM has #08 Review/UGC + #09 Winback; HF doesn't yet. NS doesn't either. Add later when there's enough customer volume to justify (currently 30 customers → review flow has thin audience).
- No Fulfilled Order metric. Backfill doesn't push order-fulfillment events; only Placed Order + Started Checkout. If a future Review/UGC flow needs Fulfilled Order trigger, add to backfill.py.
- No native 29next ↔ Klaviyo Shopify-style integration. Events come from
backfill.py(one-shot or scheduled). Ongoing sync requires either: (a) GH Actions cron like HF (memo at.github/workflows/29next-klaviyo-sync.yml), or (b) 29next webhooks pushed to a Klaviyo-events endpoint. NOT set up yet — events go stale until next backfill run.
Backfill ongoing-sync — CRITICAL OPEN ITEM
The backfill ran once (2026-05-19, --since 365d --include-carts --apply). Without ongoing sync, new orders + new started-checkouts after this run will NOT land in Klaviyo → Cart flow won't trigger on real-time abandons + segment memberships go stale.
Two paths to fix:
- (a) Periodic backfill via cron (mirror HF's GH Actions workflow at
.github/workflows/29next-klaviyo-sync.yml): runsbackfill.py --brand nailscience --since 1d --applyevery 15min. Requires GH secrets:NS_KLAVIYO_PRIVATE_KEY+NS_29NEXT_ADMIN_API_KEY. - (b) 29next webhooks pushing Placed Order + Started Checkout events to Klaviyo's events endpoint directly (real-time, no polling).
Path (a) is the proven HF pattern — recommend until/unless 29next webhook integration is wired.
Repo state (UNCOMMITTED — 12 NS/dashboard files on disk, David chose skip-git 2026-05-20)
tools/29next-to-klaviyo/backfill.py— added NS BRANDS entry (acm2.29next.store, listY4EDQu, hosttrynailscience.com).email/nailscience/bodies/01-welcome-e1.html+cart-abandon-d0..d3.html(4) +leads-d0..d3.html(4) — 9 bodies, 13-cap, FK ~7.email/nailscience/bodies/manifest.json— dashboard bodies manifest (canonical flow_ids).email/nailscience/subject-lines/manifest.json— dashboard subjects manifest (canonical flow_ids).email/dashboard/app.js— added12-leads-recoverytoFLOW_LIBRARY.email/dashboard/index.html— cache-bust?v=26.
When David does commit (his call — deferred because the branch is 1013 behind with LIVE-WAM conflicts, see Git state section): surgical git add ONLY these 12 files — NEVER git add -A (the tree also holds someone else's SkinScience bronze-chews work + hook-dirty creative/dashboard/index.html, both must stay out). NS files are all-new (no origin/main collision) so they can even land on a fresh branch off current main without the WAM rebase swamp. Klaviyo flows are already built via API (not file-deploy), so no deploy step for NS regardless.
Cross-references
humanframe_klaviyo_status.md— the original bootstrap pattern; NS mirrors it minus the bugs we then caught + fixed.klaviyo_segment_api_skill.md— authoritative segment schema; NS segments use the corrected native-metric pattern.nailscience_tracking_status.md— NS attribution (RESOLVED separately); the 29next storeacm2reference there confirmed.nailscience_klaviyo_bootstrap_brief.md— pre-bootstrap brief (written 2026-05-19; superseded by THIS status note now that bootstrap is done). Kept for the HARD RULES catalog + commerce-decision rationale.
HARD RULES re-applied (kept here for the next NS-Klaviyo session so they're at hand)
1. Liquid URL-encode filter = urlencode (no underscore). One bad filter fails the WHOLE template render silently.
2. Verify every Klaviyo template via POST /api/template-render/ — visual inspection does NOT catch render-breaking Liquid.
3. Verify segment edits by MEMBERSHIP SIZE, not HTTP 200 alone.
4. Klaviyo not-equals excludes profiles where the prop is UNSET. Use OR(existence:not-set, not-equals X) to keep the unset cohort.
5. Segment defs PATCH-able in place; flow content NOT — flow change = delete+recreate (DRAFT is safe; LIVE needs create-before-delete to not drop in-flight profiles).
6. Segment defs CANNOT reference other segments. profile-group-membership.group_ids = LIST ids only.
7. flow-start filter = exit-on-new-conversion. alltime = exclude-prior-converters. Discount/scarcity recovery flows usually want alltime (HF Cart at 83% did; NS Cart at 25% uses flow-start because moderate-discount cart should still recover returning customers re-abandoning).
8. Flow create: actions need temporary_id + links.next + top-level entry_action_id. Minutes/hours time-delays must NOT include delay_until_weekdays/delay_until_time (only days-unit).
9. Render-gate + structural parity before deleting old flow on a rebuild.
10. Honest-discount HARD RULE: every code in every send must be REAL in 29next/Shopify with a real expiry.
11. /usr/bin/curl -g for [ ] in URLs. Python stock urllib has no CA bundle → use curl.
12. Read the Klaviyo OpenAPI (raw.githubusercontent.com/klaviyo/openapi/main/openapi/stable.json) for schemas; don't brute-probe.
13. Klaviyo metrics list endpoint quirk: /metrics/?page[size]=100 may return 0; /metrics/ no-params returns all. Same for /flows/ — direct GET by id is reliable.
14. Klaviyo Client Subscribe API returns 202 even if list deleted → silent fail. Grep-audit before deleting any list.
15. Before any Netlify deploy: verify siteId via GET https://api.netlify.com/api/v1/sites/{siteId} before AND after.
16. No HF/WAM/NS env contamination: source the right ~/.klaviyo/{brand}.env explicitly; verify via /accounts/ org name before any write.
From memory nailscience_tracking_status.md
⛔ DEAD PROJECT as of 2026-06-15 (David)
NailScience is retired — do not propose or do NS work (no shadow-event rollout, no attribution changes, no maintenance). Kept for reference / template lessons only (same as BuzzerBeater). The Pattern-I / webhook history below is historical.
Status: Pattern I parity + webhook reactivation shipped (as of 2026-05-16, commit 0eb53cc)
2026-05-16 incident + fix (READ THIS FIRST)
NS lost ~1 week of Purchases (Ads Manager empty). Two root causes:
1. 29next "Order Created" webhook (store acm2.29next.store, webhook ID 67, target https://trynailscience.com/api/order-webhook, secret 1ef14a260093) was Inactive. Auto-disabled by 29next's circuit breaker after the 2026-05-11 fix's resends fired before the function deploy propagated. Inactive = 0 events / 0 errors / no event logs (deceptively "clean"). Reactivated via Update form. Re-check this webhook's Active badge regularly — it has flipped Inactive twice now.
2. Pattern I never ported to NS (deferred 2026-05-11). NS prelanders forwarded _ns_eid (wrong param name; destination reads eid_in) and NO _fbp/_fbc; syncCheckoutLinks in template/pdp/02_ecom/index.html stripped all params → 29next "No attribution found" on every order → low-EMQ Purchases. Fixed: ported HF Pattern I (commit 0eb53cc) — COOKIE_TO_URL in 3 NS prelanders (5reasons-v3, top5-advertorial index+index2) + eid_in/fbp_in/fbc_in inheritance in ns-attribution.js (canonical products/nailscience/tracking/ + 6 synced copies) + param-preserving syncCheckoutLinks in canonical template/pdp/02_ecom/index.html (overlay-mirrors to products/nailscience/index.html — do NOT edit the latter directly).
- Backfilled #101750 + #101717 via
tools/automation/backfill-29next-purchase.mjs(signed-replay tool). The May 9–15 backlog (#101652, #101586, #101520, #101488, #101396, #101355, #101257, #101256, #101060…) was NOT backfilled (David scoped to today's 2 only) — those Purchases are sunk. - RESOLVED + VERIFIED (Phase 3.5, 2026-05-16, real order #101784): quiz cross-origin path tested with a real $39.98 order. Webhook event log 1652 = 200 first attempt,
data.attribution.metadata.fb_fbc=fb.1.<ts>.PH35TEST_QUIZ_0516(literal test fbclid rode the full chain),fb_fbp/referrer(w/eid_in) populated, order Attribution UI populated (was "No attribution found"). Webhook stayed Active (no auto-pause recurrence). NailScience Meta pixel CONFIRMED =1341229411229569(distinct from HF980055147727100, no cross-brand pollution). All NS paid paths now Pattern-I-wired + deployed: trynailscience.com, /blog/five-reasons, thehealthreportdaily.com (drfranks), quiz.trynailscience.com (commits0eb53cc/c0f640c/426bf0d). Reusable verifier:tools/automation/verify-ns-attribution.mjs. Saga closed. - Standing watch item: webhook 67 (
acm2, secret1ef14a260093) auto-flipped Inactive twice historically — spot-check the Active badge every ~1-2 weeks (Phase 3.5 / any NS attribution work doubles as the check).
2026-05-19 amendment — Purchase over-count audit (commit c88b6a01)
David reported: "Purchase pixel firing more times than orders" across both brands. Audit found two in-code over-count vectors on NS (mirror of HF):
products/nailscience/checkout/index.htmlhad a nakednew Image().src=...&ev=Purchaseafter thetrackEvent('purchase','Purchase',{event_id:'purchase_'+orderId,...})call. Image had NOeid→ un-dedupable against trackEvent fbq Purchase or server webhook → every credit-card checkout double-counted. REMOVED.products/nailscience/special-offer/vip/index.htmlhad the legacy 29next funnel-templateevent:'transaction'dataLayer push. Textbook GTM Purchase-tag trigger name. RENAMED toevent:'upsell_accept'(payload preserved for any GA4 ecom listener, trigger name now unambiguous upsell signal). Bundle + Upgrade upsells did NOT have this push — only VIP.custom:UpsellViewremains the canonical NS upsell-conversion signal.- Vector 3 verified clean (2026-05-19):
acm2.29next.storehas NO Meta Pixel app installed — only thecampaignsapp. Webhook list =Campaigns App Webhook(34, 29next internal) +Order Created(67, OURS). No 29next-native CAPI ever competed with ours on NS. So the "Purchase > orders" symptom on NS was entirely Vectors 1+2 (now fixed). HF was the brand with the issue (Meta Pixel app pointed at stale pixel1953793458501776, uninstalled same day).
2026-05-19 evening — first-touch attribution capture (commit f8aacf7a)
Order #102147 (Sheri Fairbairn, $39.98, Visa) shipped "No attribution found" — investigated to a brand-new customer with a 3-week-old _fbp from a prior untracked visit, returned organic-to-apex today. Pattern I correctly had nothing to forward. To widen attribution coverage for the next return-organic / branded-search customer, added first-touch persistence + injection:
*-attribution.js: new_ns_referrercookie (set-once on first touch withdocument.referrer, 30d TTL) alongside the existing_ns_utms(last-tagged-touch within 30d) +_ns_landing(set-once).injectHiddenFields()in checkout HTMLs (NS + HF + HF echeckout): now also injects parsed UTMs (5 fields) + landing + referrer as note_attributes.order-webhook.mjs: extracts them fromnote_attributes._ns_*and ships as Meta CAPIcustom_data(utm_source,utm_medium,utm_campaign,utm_term,utm_content,landing_url,referrer).
Doesn't fix #102147 (cookies were never tagged — both her Apr 29 first touch AND today's conversion were untracked). Catches returning customers whose original visit WAS tagged + organic-search → apex entries (Google/Reddit/etc.). NS-prefixed cookies shared HF+NS — legacy namespace from HF being forked off the NS template, not worth renaming live cookies. Doctrine codified in SKILL Pattern H "First-touch capture for organic / direct-to-apex visitors" with the hard rule: _fbp timestamp older than the customer's account age = returning-organic, NOT a regression.
Full audit recipe + Vector 1/2/3 doctrine codified in SKILL Pattern H "Purchase fire-once-per-order discipline + the three over-count vectors". Grep recipe: ev=Purchase (must be 0 hits), 'event': *'transaction' (must be 0 hits on upsell pages), trackEvent.*purchase (each hit must carry event_id: 'purchase_'+orderId).
2026-05-19 — NS PDP DECOUPLED from template/pdp/02_ecom (David directive)
The compliance-footer rollout from 2026-05-14 (commit 1294e21, 7 legal pages + PDP footer wiring + Netlify/Vercel redirects) silently broke on 2026-05-16 when commit 0eb53cc (the Pattern I attribution port) re-copied template/pdp/02_ecom/index.html → products/nailscience/index.html. My footer wiring + link-color CSS lived only in the overlay; the canonical was never updated, so the re-copy restored the placeholder <li>All Products</li> / <li>Privacy</li> / etc. items. Live site bled the broken footer for ~3 days before David caught it on 2026-05-19.
David's call (2026-05-19): "I don't like that NS is coming from the template. It should be coming from its own folder."
Resolution: products/nailscience/index.html is now the NS PDP source of truth in its own folder. The template/overlay relationship is broken. Edit NS directly. Never cp template/pdp/02_ecom/index.html products/nailscience/ — it will clobber NS-specific work (compliance footer wiring, link-color CSS, ?pkg= handoff hardening, Pattern I attribution, sentinel comments). The earlier 2026-05-16 amendment line — "overlay-mirrors to products/nailscience/index.html — do NOT edit the latter directly" — is now REVERSED. The template stays as a clean starter for future brand launches; NS evolves independently. A sentinel HTML comment <!-- NS-CANONICAL: ... Do NOT regenerate from template/pdp/02_ecom/index.html --> lives above the footer block as an in-file reminder.
This decoupling is documented in skills/base-template-1/SKILL.md "Decoupled instances" section and in skills/compliance-footer/SKILL.md "Gotchas" → "Template/overlay clobbering (the NS trap)".
Status (historical): Full attribution stack live across all 3 NS surfaces, PageView dedup fix shipped (as of 2026-04-17)
NailScience prelander map (IMPORTANT — do not confuse with prelanders/nailscience/)
The NailScience funnel has two live prelanders living at the top level under prelanders/:
prelanders/drfranks/→ thehealthreportdaily.com (Netlify sitedreamy-licorice-eedba8, site_ida2bbc9e1-cf74-4df9-937d-2336b8b140fa). Dr. Franks advertorial angle. FiresViewContent+custom:prelander_cta_click.prelanders/quiz/→ quiz.trynailscience.com (Netlify sitewonderful-meerkat-49cae2, site_idb7ddc126-97b7-44a0-a457-85a772de5982). Quiz funnel. FiresViewContent+custom:quiz_started/quiz_question_viewed/quiz_question_answered/quiz_option_toggled/quiz_completed.- Both use GTM-P9MSJBP7 (different from the main NailScience container) and
../../shared/functionsfor CAPI. - Neither uses the BuzzerBeater-style
custom:prelander_viewsignal — they fire the standardViewContentevent instead. If unique prelander-level audiences are needed, port the BB pattern over. prelanders/nailscience/01/exists but is DORMANT — no tracking, no Netlify link, ignore unless being revived.
CRITICAL DEPLOY GOTCHA — missing .netlify/state.json
Both NS prelander dirs historically had NO .netlify/state.json file. Running netlify deploy --prod from an unlinked dir falls back to whatever site is globally linked in the CLI, which can push prelander content to the wrong site (happened 2026-04-15 — both prelander deploys landed on shoulder-rejuvacare.netlify.app, shoulder had to be rolled back via netlify api restoreSiteDeploy).
Always deploy prelanders with explicit --site=<site-id> flag:
```
netlify deploy --prod --dir=. --site=a2bbc9e1-cf74-4df9-937d-2336b8b140fa --functions=../../shared/functions # drfranks
netlify deploy --prod --dir=. --site=b7ddc126-97b7-44a0-a457-85a772de5982 --functions=../../shared/functions # quiz
```
.netlify/state.json files are now written in both dirs (gitignored, but persist locally). Still, prefer the explicit --site flag as belt-and-suspenders.
custom: CAPI dedup fix (commits e8bcd33 + 89dc43a, 2026-04-15)
The trackEvent() helper sends custom:<name> to fbq('trackCustom', <name>, ...) which strips the prefix client-side, but the CAPI payload was passing custom:<name> through literally — Meta dedup matches on (event_name, event_id) so every custom event was double-counting. Fix strips the custom: prefix in the CAPI payload builder (~line 287 of ns-attribution.js). Shipped to all 3 NS copies + both BB prelander copies (bb-attribution.js had the same bug, fixed in 788cc24).
Earlier status (2026-04-14)
GTM container Version 2 published live with full GA4 measurement stack. Real IDs are in place (GTM-P9MSJBP7, G-HND2VE2X2S, FB Pixel 1341229411229569). GA4 GTM Setup skill created. Persistent device-level external_id deployed across all 3 NailScience attribution.js files (2026-04-14).
Why: GA4 is the measurement backbone for performance marketing — needed before scaling paid ads. The external_id layer adds ~19% median EMQ lift per Meta's own diagnostic, and is required to hit the 8.0+ EMQ target on Purchase events.
How to apply: Use the GA4 GTM Setup skill for new product launches (measurement plane) and the new Attribution Stack & EMQ Boost skill for the JS runtime layer (skills/attribution-stack/SKILL.md).
What's live
- GTM container
GTM-P9MSJBP7— Version 2 published 2026-04-13 - 5 tags: GA4 - Configuration + 4 GA4 Event tags (Ecommerce, Lead & Shipping, Engagement, CRO Signals)
- 4 Custom Event triggers with regex matching
- 13 Data Layer Variables
- GA4 property
G-HND2VE2X2S ns-attribution.jswith persistent_ns_eidexternal_id (cookie 90d + localStorage fallback, SHA-256 hashed, passed tofbq('init')for advanced matching AND every CAPI event). Deployed 2026-04-14 to: -products/nailscience/tracking/ns-attribution.js-prelanders/drfranks/tracking/ns-attribution.js-prelanders/quiz/tracking/ns-attribution.js- All 3 files kept identical viacpfrom the nailscience source.
Still pending
1. GA4 property config — mark key events as conversions (begin_checkout, add_payment_info, generate_lead, purchase, view_item), create 4 custom audiences, link Google Ads, review enhanced measurement
2. GTM Preview mode testing — verify events fire on PDP and checkout
3. Monitor NS pixel Event ID coverage climb — 2026-04-17 PageView dedup fix (commit 493883e) deployed; Meta Events Manager → Data Sources → Deduplication keys should climb 49% → ≥75% over 24-48h rolling
4. Validate external_id end-to-end — Pixel Helper should show external_id in advanced matching, CAPI requests should send 64-char hex external_id (NOT raw ns_<uuid>), Events Manager → Test Events should show external_id contributing to EMQ
PageView dedup fix (2026-04-17)
Same bug as BuzzerBeater (commit 4fbf9b8): raw fbq('track','PageView') in every HTML snippet after fbq('init') fired browser PageViews WITHOUT eventID — Meta had nothing to match against the server CAPI PageView, so Event ID coverage was stuck at ~49%.
Fix: removed raw call from all 7 NS HTML snippets (PDP, checkout, thank-you, special-offer/bundle, special-offer/vip, drfranks prelander, quiz prelander) and added Universal PageView block at end of ns-attribution.js. Thank-you page skipped via NS.pageType === 'thank_you' guard (GTM fires Purchase there, no Pixel base snippet).
Shipped to all 3 ns-attribution.js copies byte-identical (MD5 9a7061f0b2d82cc6e54a14a41656f566):
products/nailscience/tracking/ns-attribution.jsprelanders/drfranks/tracking/ns-attribution.jsprelanders/quiz/tracking/ns-attribution.js
Deployed to 3 NS Netlify sites. Verified live — 0 raw fbq('track','PageView') on any production URL; /api/capi returns 204 on trynailscience.com.
CAPI env vars (confirmed set 2026-04-17)
The CAPI function auto-activates providers based on env vars. All 4 are set on trynailscience.com:
FB_PIXEL_ID=1341229411229569FB_CAPI_ACCESS_TOKEN= setGA4_MEASUREMENT_ID=G-HND2VE2X2SGA4_API_SECRET= set
POST /api/capi returns 204 with successful relay to both Meta and GA4.
Critical 29next SDK details (researched from SDK source)
- Purchase hook:
window.nextReady.push(function(sdk) { sdk.on('order:completed', fn) }) - SDK redirects IMMEDIATELY after order:completed —
sendBeaconis the only method that survives
Critical CAPI gotcha (also documented in attribution-stack skill)
- The CAPI function (
shared/functions/capi.mjs) does NOT hashexternal_id— it forwards whatever the client sends. The client MUST pre-hash with SHA-256. Other PII fields (em, ph, fn, ln, ct, st, zp) ARE hashed in the function — external_id is the exception.