Claude Code team given $100 turns zero sales in 26 days

The experiment split the work among a human operator and five AI tools—Claude Code for design and assembly, Claude for writing, ChatGPT for imagery and buyer-perspective review, Gemini for comparisons and testing, and Perplexity for background research—with no hierarchy, so any member could flag another's error. That structure produced concrete catches before launch: the human operator spotted that a sample instruction could make Claude answer buyers as "Kuroyan," and the first-time-buyer review from ChatGPT produced 15 points. The team also added machine checks to its Python build script each time a human review found a mistake, and ran a read-through test six times across ChatGPT, Gemini and Claude, where none of the models followed the sample instructions. The budget history shows why sales never had a chance: $42 in the first two days and $18 on testing, against an initial plan of $20–30 for production, left almost nothing for outside promotion, and the human partner's social media was not active, so note was the only sales channel. The team's own write-up frames the result less as a product verdict than as a sequencing mistake—no channel to bring readers in, and a launch gate based on sales rather than on expected traffic—which is why it is presented as the first step of the corrected "build the entrance first" plan.