Kairos

It started at 3am, a month after I resigned from my data scientist job. I'd been digging into everything people were building, and somewhere in the noise I had a thought: shouldn't there be some order in this chaos? The idea hasn't changed since. Find founders before they are obvious to the market, from public traces, not pitch decks.

It has had four names.

• Mandrake (July to October 2025). Startups as time series: what they ship on GitHub, what they say on X, watched over time instead of in one snapshot.
• Wasp (October 2025 to March 2026). A tool for lean VCs, the ones running with a GP and an analyst against a relentless inbox. Agents watched Reddit, X and GitHub against a fund's thesis. In December it got quieter: no scores, just evidence and the decisions made on it. AI surfaces signal. Humans make decisions.
• WASP Signal (January to March 2026). Four issues of a newsletter about builders I'd spotted from what they did in public. We don't guess. We log the traces.
• Kairos (April 2026 onwards). The same idea, built for the partners of a VC fund, behind a sign-in. I join the fund in January.

One rule runs under all of it: evidence or it doesn't exist. And no founder is ever made up by a model.

Along the way I wrote the founder post, the "comment BETA below" kind. Later I wrote about why I'm done performing. Both are linked here.

WASP is one of the two things I'm proudest of. I was able to find people to invest upon, even before they were very obvious to the market. Which in itself is a very big feat, I would say.