A small set of essays for founders who would rather solve a real problem than chase a thesis. Written from the evidence base of 200+ scored AI startup ideas.
Patterns from the directory: which categories carry the highest opportunity scores, which trends are over-saturated, and the five bets I'd take if I were starting fresh today.
The repeatable method behind our 200+ problem corpus: where to look, how to filter, how to score, and the trap of building from your own assumptions.
v0, Lovable and Bolt ship a working demo in an hour. Two weeks later, the codebase is unmaintainable. Here's why, and the cleanup pattern that works.
The MCP registry is a mess. Here are the eleven servers we actually use day-to-day, with the trust signals we used to pick them.
The four ways an agent silently spirals into a four-figure invoice. The guardrails that stop it. The dashboards worth checking weekly.
The 5-axis rubric we use to score every problem in the directory, and how a solo founder can run it in a weekend before committing a single sprint.