"{c.summary}"
Sharpen · Pivot · Rethink. A fair sample of what the engine produces. A page of three Pursue verdicts would read like flattery. The Rethink is the credibility lever.
Tutorial overlays walk you through each block. Click through, skip, or dismiss to explore freely.
{c.cls === 'kill' ? 'A Rethink verdict means the engine could not find an angle that survives all three skeptics. Continue reading for the Pivot Map.' : c.cls === 'pivot' ? 'The wedge is wrong; the underlying customer love is real. Pivot Map proposes 2 adjacent angles.' : 'Sharpen. The idea survives, but one assumption needs to be true. Verify before raising.'}
The one belief that, if wrong, breaks the verdict. Every memo names it. Open the Assumption Ledger for the full list.
Three structural skeptics, each looking for what would break the idea. Read the Report for the full objections and what each one needs in response.
An AI researcher who got tired of watching good ideas die in planning docs. Built PyTorch Tabular. Wrote a Packt-published book that ML teams at Amazon and Microsoft use. Now applying that same rigor to the part of building startups that usually gets done badly.
"I spent 15 years building ML systems at scale, work that lived or died on rigor. The rigor production ML demands rarely makes it into how founders evaluate their own ideas.
I kept watching good ideas die in planning docs and bad ideas survive long enough to consume two years of someone's life. The validation tools I tried either flattered every idea or summarized a market without ever finding the assumption that would kill it. So I built the system I wished I'd had. Adversarial by design. Structured to produce an artifact a founder could defend."
{p.b}