AICofounder

Inspiration

Every founder has a graveyard of half-baked ideas and no cheap way to know which one is worth quitting your job for. Validation usually means weeks of building, cold outreach, and guesswork. We wanted a cofounder in your pocket — one that turns "I think people want X" into actual signal before you write a single line of product code.

What it does

AICofounder takes a startup idea described in one sentence and:

• Extracts your target customer, the core problem, and your differentiator
• Generates a real, conversion-focused landing page from that idea
• Publishes it at a shareable link and collects email signups
• Runs built-in A/B testing on headlines and framing
• Rolls everything into a transparent 0–100 validation score so you know if the idea has legs

How we built it

A Vite + React + TypeScript frontend with a Tailwind UI, backed by an Express server that is the only process that talks to the LLM — the browser never touches an API key. We use Claude for idea extraction and copy generation, and Perplexity for live competitor and market research. The whole thing runs offline with stubbed responses too, so demos never depend on a flaky network.

Challenges we ran into

Keeping the LLM output structured and reliable enough to drive a real UI was the hard part — free-form model responses don't map cleanly onto a scoring rubric. We also had to design the validation score so it felt honest instead of a vanity number, and wire up A/B testing without turning the codebase into spaghetti.

Accomplishments that we're proud of

A full idea → landing page → signups → score loop that actually works end to end, and a fallback stub mode so the entire product demos with zero keys. The scoring is transparent — you can see why an idea got the number it got.

What we learned

That the gap between "an idea" and "something you can test" is mostly plumbing, and that plumbing can be automated. We also got a lot better at forcing LLM output into strict schemas without losing the useful nuance.

What's next for AICofounder

Deeper market research, automated ad spend to drive real traffic to generated pages, cohort tracking over time, and a "pivot suggester" that reads your signups and recommends the next wedge to test.

Built With

• react
• tailwind
• typescript
• vite