Venture decision system · Independent
StartupOS
Build evidence, not assumptions.
- Status
Live · private workspace
- Built with
- Next.js, TypeScript, Supabase with row-level security, Claude Code (local assistant)
A workspace where founders grade hypotheses, experiments and decisions on an explicit evidence ladder. The AI assistant proposes structured changes that the founder has to approve.
Problem
Early-stage teams talk about “validation” as if it were one thing. In practice they mix opinions, anecdotes and real evidence. Decisions get made on whichever was said most confidently.
Insight
Make the evidence explicit and graded. Validation isn't one bar. It's four (problem, solution, business, growth), each rated on a ladder from no evidence to strong. An AI assistant should be held to the same rule as humans: it can propose, but the founder decides.
What I built
A venture workspace following the chain opportunity → hypothesis → assumptions → experiments → evidence → decisions → next action. Every validation level is backed by supporting and contradicting evidence and a written rationale.
The built-in assistant captures evidence and tasks directly. Anything that is a judgment, such as a validation level, a rating or a stage change, arrives as a proposal the founder can apply or dismiss. There is no delete action at all.
My role
Concept, methodology, product design and build.
Evidence
- Live as a private, invite-only workspace
- 136 automated tests
- Per-venture access control enforced in the database
What it shows
- Product design grounded in entrepreneurship method
- Designing AI assistance with clear authority boundaries