The territory
A backlog of AI product ideas with no way to rank them except conviction. Conviction is not a prioritisation method — it just means the loudest idea ships first.
What I did
Built twelve prototypes in six weeks, each scoped to answer one specific question rather than to look finished. Most were disposable by design — the prototype was the question, not a preview of the product.
Where it stands
Three of the twelve questions turned out to matter. Those three got funded; the other nine are documented as closed, not abandoned.