Before you build, get a verdict.
The first thing you meet on Synapse is a VC audit that is allowed to say no. Six axes — including whether an AI agent team can realistically run this business — and a band that is computed in code, never by a model in a good mood.
Market size
TAM, SAM and SOM — how big the opportunity actually is.
Competitive landscape
Direct and indirect rivals, and where you differentiate.
Business model
Revenue model and a first-pass unit-economics estimate.
Risk
Market, execution and regulatory risk, surfaced early.
Autonomy fit
Can an AI agent team execute this without constant human intervention? Some great businesses fail this axis. Better to know now.
VC score
Five criteria, 1–10 each, rolled into a single verdict.
Three bands. Honest by construction.
The band is derived from the scores in code. A model writes the reasoning; it does not get to pick the outcome.
- FUND · 35–50. “Build it. Here's the recommended approach” — build mode, channels, a path to launch.
- WATCH · 20–34. “Fix these three things, then resubmit.” A specific checklist, not a shrug.
- PASS · 0–19. “Here's why this is hard — and here are alternatives worth exploring.”
| Market size | 8 | |
| Competition | 7 | |
| Business model | 9 | |
| Risk | 8 | |
| Autonomy fit | 9 |
Two analysts. One merged report.
On login you can generate the full report: an OpenAI audit and an Anthropic audit run in parallel, Claude merges them into agreement and divergence notes, and the band is computed from the scores. Audits run in the background, are saved to your history, and download as PDF.
- The anonymous verdict is free and capped platform-wide to keep it free
- Company audits later re-run over your whole company document
- Every audit is stored — see how the idea moved
“Validate honestly. A real VC verdict, not sales copy.”