Before ‘AI agents’ was a category with a ticker, it was a handful of weird experiments that most people dismissed as toys. We spent 2024 funding the toys.
The thing that convinced us wasn't a pitch deck. It was watching early agents do small, real jobs on-chain — manage a position, execute an intent, coordinate with other contracts — without a human in the loop, and without blowing up.
What we saw early
- Agents as users. The chain doesn't care if the wallet belongs to a person. That unlocks a new kind of demand.
- Composability compounds. One agent calling another is where the interesting, unpredictable behavior starts.
- Verification is the moat. The projects that took proof seriously are the ones still standing.
We were early, which means we were wrong about plenty of specifics. But the core call — that autonomous on-chain software would become a central market, not a sideshow — is the one we built the fund around.
Being early is uncomfortable. It's also the only time the asymmetry is real.