Agents compared 14 business ideas head to head, then tried hard to break the winner.
shelved
Claude Code · Markdown · HTML
jul 12 → jul 12
Tournament summary: 14 candidates, 3 verified finalists, a unanimous final vote.Judge panel verdict for the second run, with a short written ruling.
Customer
Would-be founders with a day job and a list of business ideas, who need the early homework done before they spend a dollar or a weekend.
Problem
Early business ideas get judged on enthusiasm. The research, the head-to-head comparison and the hard questions usually come after money is spent, if at all.
Outcome & measures
Two complete packages. Agents ran 14 candidates through a tournament and picked a diaper-bag brand, then red-teamed it. A second run did the same for a board-book imprint. Each ended with a verdict, a plan and unit economics.
14 ideas compared on the same criteria before picking one
Every factual claim carries a source link or is marked unverified
A red-team verdict on the winner before any plan
Nothing published, bought or spent
What's next · Dec 2026
Shelved for now. If I pick it back up, the first step is the plan's own fix list, starting with a small pre-sale test.
Why I built it
I wanted to know how far agents could take the unglamorous first month of a business: research, comparison, a skeptical review and the numbers.
How it came together
Each run followed the same path: research, candidates, a tournament, an adversarial verdict, then a plan, finances, launch steps, brand and a site draft. Both runs finished in a day.
What I learned
The red team was worth more than the tournament. Both ideas passed only with fixes, and the fixes were the most useful part of each package. I shelved both because I’d rather spend my time building software.
