Evidence Gate

Every claim in your inbound deal flow, graded by evidence.

An AI analyst that reads each application and tells you which claims are backed by artifacts inside the pack, which rest on the founder's word, and which cannot be checked as presented — before your team spends an hour on it.

Send applications from your current batch Not investment, legal, or financial advice. The gate grades how verifiable claims are as presented — it does not verify whether a claim is true. Humans make the decisions.

AI made starting a company cheap. Your inbox knows it.

Application volume grows; screening capacity doesn't. Decks are increasingly AI-written: specific-sounding numbers, polished narratives, round metrics. Summarization tools read them faster — but a faster summary of an unverifiable claim is still an unverifiable claim. The question that costs your team the most time is not "what does this startup do" but "what here can I actually check?"

One brief per application. Four things in it.

1. Verdict

proceed / hold / escalate. Readiness for partner review, never an investment view.

2. Claim audit

every material claim, one of three grades: supported in pack, self-reported, unverifiable as presented — plus the exact artifact that would verify it.

3. Red flags

only patterns visible in the pack: metric inconsistencies, arithmetic breaks, unnamed "major" partners, unfalsifiable statements.

4. Questions for the founder

artifact-based, answerable, ready to paste into your reply.

Try it on a sample application

Both sample applications are illustrative — the companies are not real. Each brief is the pre-generated output of a real agent run on that sample.

Replaying a recorded agent run…

How it works

  1. You send applications from your current pipeline — PDF or text, any format your intake produces.
  2. Each one goes through a claim-level evidence audit.
  3. You get one brief per application within 48 hours.

What this is not