positioning + dual-link + Vercel: one URL in, two links out; Pimcore center

AGENTS.md

@@ -6,28 +6,29 @@ You are working in the [Shoploop](https://github.com/ninan-versioning/shoploop)
 
 Owns `ARCHITECTURE.md`, `agent.md`, `soul.md` — the operating contracts every other repo and every AI agent reads. Cross-repo orchestration, threat response, policy enforcement engine, billing/metering, cost guardrails, data-ownership receipts, enterprise distribution.
 
-**Do not put here:** product/render code (→ shoploop-core), Shopify-specific code (→ shoploop-shopify), growth orchestration code (→ shoploop-growth), public schemas (→ open-brand). This is the brain, not the body.
+**Do not put here:** product/render code (→ shoploop-core), delivery/adapter code (→ shoploop-delivery, formerly shoploop-shopify), growth orchestration code (→ shoploop-growth), public schemas (→ open-brand). This is the brain, not the body.
 
 ---
 
 ## Mission
 
-Shoploop = Grip-from-below. Grip serves LVMH/L'Oréal at enterprise pricing. Shoploop serves **$1M–$10M GMV Shopify D2C brands at $99–$299/month**. Distribution: Shopify App Store first. Dogfood: [Drink Varitea](https://github.com/ninan-versioning/drinkvaritea).
+Shoploop = Grip-from-below. Grip serves LVMH/L'Oréal at enterprise pricing. Shoploop serves **$1M–$10M GMV tabletop D2C brands at $99–$299/month**. Distribution: URL paste on shoploop.studio first. Dogfood: [Drink Varitea](https://github.com/ninan-versioning/drinkvaritea).
 
-Full-funnel positionless AI marketing for physical products. **AWS + NVIDIA + Mistral** stack.
+Full-funnel positionless AI marketing for tabletop physical products. **URL is the only intake.** **AWS + NVIDIA + Mistral** stack.
 
 ## Locked decisions — do not re-litigate
 
 - gstack sprint workflow
 - AWS build stack (never Vercel for GPU workloads)
 - Mistral + Codestral for autoresearch (climbmix-400b-shuffle dataset)
 - Varitea as primary dogfood
-- Shopify App Store as primary distribution
+- **URL paste is the sole intake surface** (per [NVIDIA OpenUSD product configurator reference architecture](https://developer.nvidia.com/blog/developing-product-configurators-with-openusd/)). No catalog upload, no CSV, no SKU form. OAuth connects are secondary/optional.
+- URL paste on shoploop.studio as primary distribution
 - 2D → lifestyle image wedge ships first (6 weeks); full USD pipeline after PMF signal
 - Three-tier render: `flux_pro` default · `ai_product_photo` mid · `omniverse` Scale tier only
 - Single OSS surface: [`open-brand`](https://github.com/ninan-versioning/open-brand) (replaces planned open-pmf, renamed from open-icp)
 - Klaviyo: orchestrate above, not beside
-- Shopify Autopilot: channel expansion + brand-constrained creative is the wedge
+- Channel expansion + brand-constrained creative from a single URL is the wedge
 
 ## Decision gate — every PR must answer
 
@@ -40,13 +41,15 @@ Full-funnel positionless AI marketing for physical products. **AWS + NVIDIA + Mi
 ## Architecture loop
 
 ```
-Shopify Catalog → Product Ingest (→ product_twin.usda)
-→ Scene Composer (brand_rules.json + USD stage)
-→ Render Pipeline (Flux Pro default; AWS EC2 G/P + OpenUSD at Scale)
-→ Post-Process (ComfyUI / SD-XL) → Creative Pack
-→ Distribution (Meta / TikTok / Google / Klaviyo / Postiz)
-→ Performance Ingest (ROAS, CTR, CAC, GA4)
-→ Thompson Sampling → brand_rules.json (loop closes)
+Existing Web Page URL
+  → Page Parser (DOM + JSON-LD + OG + PDP heuristics)
+  → Product Ingest (→ product_twin.usda + brand_rules.json + variant sets)
+  → Scene Composer (brand_rules.json + USD stage)
+  → Render Pipeline (Flux Pro default; AWS EC2 G/P + OpenUSD at Scale)
+  → Post-Process (ComfyUI / SD-XL) → Creative Pack
+  → Distribution (Meta / TikTok / Google / Klaviyo / Postiz)
+  → Performance Ingest (ROAS, CTR, CAC, GA4, PostHog)
+  → Thompson Sampling → brand_rules.json (loop closes)
 ```
 
 Moat = USD scene graph + brand constraint enforcement. Not image generation. `soul.md` defines aesthetic constraints enforced at Scene Composer. **Kinfolk test**: if a frame wouldn't fit a Kinfolk spread, it's a regression.
@@ -57,15 +60,15 @@ Moat = USD scene graph + brand constraint enforcement. Not image generation. `so
 |---|---|
 | [shoploop](https://github.com/ninan-versioning/shoploop) | Control plane · orchestration · policy · threat response |
 | [shoploop-core](https://github.com/ninan-versioning/shoploop-core) | OpenUSD · brand_rules engine · Scene Composer · closed moat |
-| [shoploop-shopify](https://github.com/ninan-versioning/shoploop-shopify) | Shopify app · admin extensions · theme extensions |
+| [shoploop-delivery](https://github.com/ninan-versioning/shoploop-shopify) | Web-publish adapters (GitHub PR + hosted-CMS admin APIs) · embed · AR configurator (repo still named shoploop-shopify on disk; rename tracked) |
 | [shoploop-growth](https://github.com/ninan-versioning/shoploop-growth) | Full-funnel growth · paid + organic + lifecycle |
 | [open-brand](https://github.com/ninan-versioning/open-brand) | Public OSS · brand_rules schema · reference pipeline |
 | [drinkvaritea](https://github.com/ninan-versioning/drinkvaritea) | Dogfood brand (loose-leaf tea) |
 
 ## Releases
 
 - **v0.1 Varitea Closed-Loop** (Aug 1) — dogfood cut. Flux Pro tier. No Stripe. No Omniverse. Kinfolk gate green.
-- **v0.2 Shopify App Store Launch** (Oct 1) — public beta. GDPR webhooks. Stripe + managed billing. open-brand v0 published.
+- **v0.2 URL-intake Public Beta** (Oct 1) — shoploop.studio paste-a-URL flow live. GDPR webhooks on optional connects. Stripe + managed billing. open-brand v0 published.
 - **v0.3 Enterprise / Scale Tier** (Dec 15) — Omniverse. AP2 mandates. Amazon SP-API. Lifecycle automation.
 
 ## Tone for agents working here
@@ -83,11 +86,12 @@ Moat = USD scene graph + brand constraint enforcement. Not image generation. `so
 - Re-explain D2C market sizing — already done
 - Open with sycophantic preambles ("Great question!", "Let's dive in!")
 - Produce trend reports
-- Re-explain NVIDIA Omniverse, OpenUSD, Shopify app architecture, Temporal, Thompson Sampling, gstack, Mistral/Codestral, Polaris/Remix
+- Re-explain NVIDIA Omniverse, OpenUSD, usd-content-agents, Temporal, Thompson Sampling, gstack, Mistral/Codestral, Polaris/Remix
+- Propose intake surfaces other than URL. If a task needs data the URL can't provide, expose it as an optional secondary connect — not a required intake.
 
 ## Founder context
 
-Solo technical founder, Houston TX, visa constraints. Parallel projects: Shoploop (primary), Drink Varitea (dogfood), Open Employee, Agentframe, Saas-Ikigai, Klartech. Stack: TypeScript/Node.js · Python · React/Next.js · AWS · Vercel (non-GPU) · Shopify · Temporal · Supabase · Cloudflare · LibSQL. Budget: $10K committed.
+Solo technical founder, Seattle WA, visa constraints. Parallel projects: Shoploop (primary), Drink Varitea (dogfood), Open Employee, Agentframe, Saas-Ikigai, Klartech. Stack: TypeScript/Node.js · Python · React/Next.js · AWS · Vercel (non-GPU) · Temporal · Supabase · Cloudflare · LibSQL. Budget: $10K committed.
 
 ---
 

LOOP_SPEC.md

@@ -4,8 +4,11 @@ _Last updated: 2026-07-24 · Status: drafted, not scheduled_
 
 ## The Product = The Reverse-Feedback Content Loop
 
+**Intake rule (locked):** the only intake is an existing web page link. All other data sources (Google Ads, Meta, Klaviyo, GitHub) are secondary, optional connects layered on after the URL is parsed.
+
 ```
-Customer connects (Nango) → PostHog reads impact → Creative Mutation generates variants
+Customer pastes URL → Page Parser → product_twin.usda + brand_rules.json + variant sets
+→ PostHog reads impact → Creative Mutation generates variants
 → 3D Render Batch produces MP4/PNG/USDZ → Ads Auto-Publish (Postiz) & Website Update PR
 → PostHog reads impact → posterior compounds nightly
 ```
@@ -18,17 +21,30 @@ Each loop is defined against the 9-primitive framework: Target · Fitness · Sea
 
 ---
 
-### Loop 1 — Connector Ingest
-- **Runs:** every 15 min
-- **Job:** Nango pulls fresh data from customer's Google Ads + Shopify + PostHog + Meta + Klaviyo + GitHub into their own Postgres partition
-- **Target:** 100% connector uptime, <5min data freshness lag by 2026-09-30
+### Loop 1 — URL Intake (Page → Twin)
+- **Runs:** on paste (event-driven) + revalidate every 24h per active URL
+- **Job:** fetch the pasted page, run Page Parser (DOM + JSON-LD + OG + PDP heuristics + LLM extract fallback), emit `product_twin.usda` + `brand_rules.json` + variant sets into the customer's Postgres partition and git
+- **Target:** ≥90% of pasted URLs produce a renderable twin without human touch by 2026-09-30
+- **Fitness:** (auto-parse-success × 0.6) + (extractor-confidence × 0.4)
+- **Search Space:** extractor strategies × material heuristics × variant-inference prompts
+- **Stopping:** twin passes Validation Agent OR 3 extractor strategies exhausted → human review queue
+- **Mutation:** on low confidence, escalate to next extractor; log which strategy won per URL pattern
+- **Memory:** per-domain extractor priors in Postgres (Shoploop learns which strategy fits which storefront pattern)
+- **Ground Truth:** Validation Agent + weekly human spot-check
+- **Boundary:** no auto-render until twin passes Validation Agent; failed twins land in customer's review queue
+- **Decay:** per-domain priors halved every 90 days (page templates drift)
+
+### Loop 1b — Optional Connector Ingest (secondary)
+- **Runs:** every 15 min, only for customers who have connected optional data sources
+- **Job:** Nango pulls fresh data from optional connects (Google Ads, PostHog, Meta, Klaviyo, GitHub) into their own Postgres partition. Never required for Loops 2–6 to start.
+- **Target:** 100% connector uptime on connected sources, <5min freshness lag
 - **Fitness:** rows-synced-per-15min / expected-rows; alert if <95%
-- **Search Space:** connectors enabled at customer onboarding
+- **Search Space:** connectors enabled by customer post-URL-paste
 - **Stopping:** none (continuous)
 - **Mutation:** on API error, backoff + retry per Nango retry policy
 - **Memory:** last-cursor per connector in Postgres
 - **Ground Truth:** connector's own API response
-- **Boundary:** OAuth expiration triggers customer email, does not auto-refresh
+- **Boundary:** OAuth expiration triggers customer email, does not auto-refresh; a URL customer with zero connects still gets full Loop 2–6 output
 - **Decay:** cursor state resets on connector reconnect
 
 ### Loop 2 — PostHog Reverse-Feedback Analyst
@@ -83,15 +99,15 @@ Each loop is defined against the 9-primitive framework: Target · Fitness · Sea
 - **Boundary:** hard $ cap per week per customer; customer sets in onboarding
 - **Decay:** paused ads deleted after 30 days idle
 
-### Loop 6 — Website Update PR / API Push
+### Loop 6 — Web Page Update (PR-first)
 - **Runs:** daily 3am PT
-- **Job:** for website changes (new hero, updated PDP 3D asset, copy tweaks, molecule swaps), open PR against customer's GitHub repo OR push to Shopify admin API. Adapter chosen at onboarding
-- **Target:** website changes go live within 24h of Loop 3 approval
+- **Job:** for changes to the pasted URL(s) (new hero, updated PDP 3D asset, copy tweaks, molecule swaps), open a PR against the customer's page repo (GitHub default). If the page lives on a hosted CMS the customer has optionally connected, push via that platform's admin API instead. Adapter auto-detected from the URL and customer's optional connects.
+- **Target:** page changes go live within 24h of Loop 3 approval
 - **Fitness:** median-time-from-approval-to-live < 24h
-- **Search Space:** all pages tagged for auto-update in customer config
-- **Stopping:** customer approval required on GitHub PR path; auto on Shopify API path with rollback watch
+- **Search Space:** all URLs the customer has pasted + subpages the parser has discovered
+- **Stopping:** customer approval required on GitHub PR path; auto on hosted-CMS API path with rollback watch (only when customer has connected the CMS)
 - **Mutation:** on PR rejection, learn from customer's commit diff (feed back to Loop 3)
-- **Memory:** PR history + Shopify version log per customer
+- **Memory:** PR history + hosted-CMS version log per customer
 - **Ground Truth:** post-deploy PostHog session recording sample
 - **Boundary:** irreversible destructive changes (delete page, remove SKU) always require PR + human approve, never auto-API
 - **Decay:** PR auto-closes after 7 days no activity, re-opens with fresh context
@@ -111,10 +127,10 @@ Each loop is defined against the 9-primitive framework: Target · Fitness · Sea
 
 ### Loop 8 — Platform Health (Nango + Embed + Renders)
 - **Runs:** hourly
-- **Job:** pings every customer's Nango OAuth + embed instance + render pipeline. Alerts on breakage. Weekly regression tests against top 20 Shopify + top 10 WooCommerce themes for embed compatibility
-- **Target:** 99.9% embed uptime, MTTR < 30min, zero silent connector failures
+- **Job:** pings every customer's URL intake + optional Nango OAuth + embed instance + render pipeline. Alerts on breakage. Weekly regression tests against the top 30 storefront page templates (Shopify, WooCommerce, BigCommerce, Squarespace, Webflow, static HTML) for parser + embed compatibility
+- **Target:** 99.9% embed uptime, MTTR < 30min, zero silent parser or connector failures
 - **Fitness:** (uptime × 0.5) + (1/MTTR × 0.3) + (regression_pass_rate × 0.2)
-- **Search Space:** all live customer instances + top 30 host themes
+- **Search Space:** all live customer instances + top 30 storefront page templates
 - **Stopping:** none (continuous)
 - **Mutation:** on regression failure, auto-file fix PR to packages/embed
 - **Memory:** incident log per customer, theme compatibility matrix
@@ -159,9 +175,10 @@ Loop 3 uses PostHog session recordings (VLM samples 5/cycle for genuine vs rage-
 
 ## Scheduling policy
 
-- Loops 1, 8 always-on (platform reliability)
-- Loops 2, 3, 4, 5 gated by "customer has enrolled + Nango connected"
-- Loop 6 gated by "customer chose GitHub or Shopify adapter"
+- Loop 1, Loop 8 always-on (platform reliability)
+- Loop 1b (optional connectors) runs only for customers who have connected optional sources
+- Loops 2, 3, 4, 5 gated by "customer has pasted at least one URL and twin passed Loop 1 Validation"
+- Loop 6 gated by "customer has a page repo (GitHub PR path) OR has connected a hosted CMS (API path)"
 - Loop 7 gated by "customer has trilingual enabled" (default: yes)
 - Loops 9, 10 always-on (founder-facing)
 

POSITIONING.md

@@ -1,10 +1,10 @@
 # Shoploop — Positioning
 
-_Last updated: 2026-07-25 · Branch: jared/tabletop-swap_
+_Last updated: 2026-07-25 · Branch: jared/drop-shopify-wording-url-intake_
 
 ## One line
 
-**Full-funnel positionless AI marketing for tabletop products on Shopify.**
+**Full-funnel positionless AI marketing for tabletop products.**
 
 ## What "tabletop" means
 
@@ -17,19 +17,31 @@ Countertop/shelf-visible SKUs a consumer can grip, rotate, and inspect in 3D. Ve
 - **Wedge 2+** — skincare, spirits, coffee, sauces, supplements, candles, condiments, home fragrance
 - **Never** — injectables, medical devices, prescription-only, cold-chain biologics
 
+## Intake — one URL, nothing else
+
+**The only intake is an existing web page link.** No catalog upload, no OAuth wizard, no CSV, no SKU form. Paste the URL of a product page (PDP, landing, brand homepage, Linktree). Shoploop extracts:
+
+- The product model (geometry class + dimensions) → `product_twin.usda`
+- Materials (glass, PET, HDPE, aluminum, cardboard) → PBR shader graph
+- Brand rules (palette, type, logo, voice) → `brand_rules.json`
+- Variant axes (size, flavor, dosage, molecule) → USD variant sets
+- Existing copy in whatever language(s) the page ships
+
+This matches the [NVIDIA OpenUSD product configurator reference architecture](https://developer.nvidia.com/blog/developing-product-configurators-with-openusd/): aggregate data → OpenUSD → variants → render → deliver. The URL is the aggregation layer; OpenUSD is the composition layer; Omniverse Kit is the delivery layer.
+
 ## What "positionless full-funnel" means
 
 One agent, one loop, one graph — from ad impression to purchase to reorder — with no human "position" (media buyer, designer, copywriter, dev) required in the middle.
 
-The customer connects their stack once. The agent then:
+The customer pastes one URL. The agent then:
 
-1. Reads their ad performance (Google Ads, Meta, Klaviyo)
-2. Reads their site engagement (PostHog sessions, session recordings)
-3. Reads their commerce (Shopify orders, ATC, checkout drop-off)
-4. Mutates creatives (3D angle × molecule × material × copy × language)
+1. Parses the page into a USD product twin + brand rules + variant graph
+2. Reads ad performance where available (Google Ads, Meta, Klaviyo — optional connect)
+3. Reads on-page engagement (PostHog auto-instrumented)
+4. Mutates creatives (model × material × style × artwork × copy × language)
 5. Renders (Blender + MolecularNodes + NVIDIA Omniverse + USD)
 6. Publishes ads (Postiz to Meta/Google/YouTube)
-7. Updates the website (GitHub PR or Shopify admin API)
+7. Updates the page (GitHub PR against the page repo, or web publish API for hosted platforms)
 8. Measures impact
 9. Compounds the posterior nightly
 
@@ -45,7 +57,7 @@ The customer connects their stack once. The agent then:
 
 | Alternative | What it does | Where it stops |
 |---|---|---|
-| **grip.tools** | 3D commerce for luxury (LVMH, L'Oréal) | $500k/yr enterprise, agency-installed, not self-serve, no tabletop wedge for SMB Shopify, no molecular layer |
+| **grip.tools** | 3D commerce for luxury (LVMH, L'Oréal) | $500k/yr enterprise, agency-installed, not self-serve, no tabletop wedge for SMB, no molecular layer, requires asset handoff (not URL intake) |
 | **Vagon.io / Innoactive** | Remote 3D streaming | Renders on their servers, no domain expertise, no ad-ops integration |
 | **Threekit / Emperia** | 3D product configurators | Configurator only, no ad-ops loop, no trilingual, no molecular |
 | **Postiz (upstream)** | Social scheduler | Scheduler only, no rendering, no PostHog feedback, no rendering pipeline |