Stars are vanity. Commits are a diary. Here is how we read a technical founder's public record before we ever get on a call, and how we turn it into a score you can defend.
You are about to spend an hour with a founder. Before that hour, most investors read a deck and a LinkedIn. Both are written to impress you. The founder's public code was not. That is the point. This is one input into founder scoring, not a replacement for it.
The signal, not the star count
A repo with 4,000 stars tells you the founder can market. It does not tell you they can build. We ignore the trophy case and read the boring stuff: commit cadence, how they respond to issues, whether they refactor their own mess, and how they write to strangers who file bugs. It is the same shift recruiters sourcing engineers off GitHub in 2026 have made: proof of work over claimed work.
The resume is what someone says they did. The record is what they actually did, timestamped.
What we actually read
We score a short list of signals that correlate with founders who ship and keep shipping when it gets hard:
- Commit consistency over 18 months, not a burst before fundraising.
- How they handle a rude issue. Grace under pressure shows up in a comment thread.
- Whether they delete code. Founders who cut scope tend to survive.
- Docs written for other people, a proxy for whether they can hire and delegate.
Each signal maps to a point on your bar, and every point links back to the exact commit, thread, or file that earned it. No black box, the same discipline behind evidence-based scoring. If a partner asks "why an 8 and not a 6," the answer is one click away.
Where resumes miss
Titles inflate. A "Staff Engineer at BigCo" may have shipped one feature in three years behind four layers of review. A founder with no fancy title may have carried a product alone. The public record flattens the title and shows you the work, the same reason we put more weight on the record than the room in reference calls flatter, the record doesn't.
Bring it to the call
The goal is not to replace the conversation. It is to walk in already knowing where to push. You stop asking "tell me about your background" and start asking "walk me through why you rewrote the auth layer twice in Q3." That is a better hour, for both of you.
Your judgment still makes the call. We just make sure it walks in with the receipts.