The problem is not “Figma for the factory.” The problem is a completed DRA report with no prototype. PTG picks the highest-confidence report still unprototyped — BRI idea or GHT repo — and runs a 2-stage MiniMax pipeline: plan/requirements/screen list, then one HTML document per screen. Viewer: in-app iframes + tab strip. No download, no new tab. Hourly chain: :18 BRI→DRA, :20 DRA report, :23 GHT→DRA, :25 PTG. Pre-2026-08-17 selector was BRI-only and generated nothing while BRI was idle — now GHT counts. Prototypes are per-user private. Public landing. Marker format for long HTML (not JSON-escaped bodies). MiniMax unpaid → visible skip, not a fake app. Live generation uses the platform MiniMax chain.
Zero human input on the happy path. You still open the viewer to judge whether the HTML is garbage.
Not Figma, not v0.dev screenshots.
GHT feeder after the BRI-idle bug. Selector is confidence, not recency-only.
Static HTML. No backend. No auth inside the iframe app.
If DRA has no completed reports, PTG correctly does nothing.
Will not ship a second code generator. Landing lists no HTML bodies.
Opportunity → clickable mock. Not customer-facing SaaS.
Cards match the PTG README.
Login for /ptg/app. This page is not your prototypes.
Open PTGRequires a MelTuc account. Create one free.
DRA report → 2-stage MiniMax plan + multi-screen HTML prototype. Hourly picker. Private per user. Health stays cheap. If this page disagrees with the signed-in app, trust the app — the product contract and tests enforce that rule.