graduating the prototypes
Two colleague-built AI-Studio apps — a programmatic-ad screenshot deck builder and a CMS component-usage auditor — rebuilt onto the platform as owned systems. Not like-for-like moves.
team one internal tools, deployed in production on the agency's internal aws/eks platform. requesters and any credential history withheld.
Ad deck builder. Programmatic buys need “screenshot decks” proving placement, but no fixed URL exists for a served ad, so a now-gone offshore team spent two to three days per deck — and a planner's AI-Studio prototype couldn't be published. The audit found CSS device frames that never exported, a third-party screenshot service with bot spoofing and fabricated fallbacks, hardcoded fake slot detections, no persistence, no tests. The rebuild owns the capture: headless Chromium with honest consent handling and a labeled “representative mockup” fallback (the rule is never fake the page), DSP share-URL ingestion with server-side video hero-frame sampling (player media is CORS-blocked client-side), nearest-standard-size readouts, a server-side rasterizer for WYSIWYG decks, and an AC01–AC20 coverage matrix with only Chromium, the model, and storage mocked.
CMS component usage. An applet auditing component usage across three CMS sites talked from the browser to a database with read-and-write open to anyone and a hardcoded client-side password. Re-platformed onto Express and Postgres with server sessions, a custom export run in-pod to move the data, the fleet's rotation self-heal, and the agency design system. Both graduated in days; both are on the telemetry fleet.
Full case study in progress — artifacts pending clearance or writing time. The insight below is already earned.
A prototype that can't be published isn't a prototype problem, it's an ownership problem. Graduating it means auditing what it actually does — spoofed screenshots, hardcoded detections, open database rules — and rebuilding the honest version, not re-hosting the demo.
the insight