Week 1: From idea to conviction — Corexi takes shape
The idea behind Corexi existed for two years. This week we stabilized the product thesis through market-fit research and committed to building it.
Corexi wasn't a sudden pivot. The core idea — using AI to analyze, improve, and eventually generate better product experiences — has been in our heads for nearly two years.
The original vision
The earliest version of this idea was broader: AI-powered generative UI, automated A/B testing, and one-click deployment of winning variants. Analyze a product, generate an improved version, test it against the original, and push the winner live. A closed loop.
It was ambitious. Too ambitious for a first product, as it turned out.
What changed
Over the past months we ran deep market-fit research. Talked to product teams, design leads, and engineering managers. The pattern was consistent:
- Teams don't need AI to generate new UI. They need AI to tell them what's wrong with the UI they already shipped.
- The EU Accessibility Act (June 2025) turned WCAG compliance from "nice-to-have" to "legal requirement." Teams need a defensible audit trail, not another design tool.
- AI agents (Cursor, Claude Code, Windsurf) became the primary coding surface. Tools that don't live inside the agent lose the loop.
The broad vision — generate, test, deploy — is still where this goes long-term. But the right entry point is narrower: scan the running product, find what's broken, and ship the fix as code directly into the IDE.
This week
We committed. Stabilized the product thesis, scoped Phase 1 (Scan + Fix in your IDE), and started building. Not a prototype, not a slide deck — the actual product.
The idea is two years old. The conviction is one week old. That's the difference.
Enjoyed this? Get the weekly build log in your inbox.