Data Layer.
Data is the missing crypto-AI moat — but most data layers are still solving the wrong problem.
Thesis
Why this sector, why now.
Everyone agrees AI is data-constrained. Decentralized data layers (Grass, Vana, Masa, others) bet that crypto rails unlock data supply that proprietary platforms cannot — residential bandwidth, user-permissioned datasets, verifiable provenance. The thesis is right. The execution gap is selling data into AI labs at scale and proving quality competitive with scraped Common Crawl + licensed sources. The leaders are running ahead on user count (airdrop-driven). The signal I want is enterprise data-buy contracts, not airdrop snapshots.
Signals I track
What would move my read.
- 01
Disclosed data-licensing deals with AI labs
- 02
Post-airdrop user retention
- 03
Data verifiability primitives shipping to production
Kill shot
What would kill the thesis
Synthetic data quality crosses the bar for foundation model training, and the marginal value of human-supplied web data collapses to near-zero.
Coverage
Projects on the radar.
Going deeper
Bespoke Data Layer dive for your fund.
Draft thesis · Grey voice in progress, edits land continuously