AI Agents.
The first crypto-native category where the product is software that earns money on its own.
Thesis
Why this sector, why now.
Agents are the first AI primitive that fits crypto's settlement layer naturally. They need wallets, identity, payments, and counterparty risk priced in code — exactly the things blockchains were built for. The interesting fork is whether agent value accrues to the framework (Eliza, Virtuals), the agent itself (token-as-equity), or the substrate that pays for agent work. My read: framework is undifferentiated long-term; substrate is winner-take-most; individual agent tokens are mostly noise with a few signal exceptions.
Signals I track
What would move my read.
- 01
Agent revenue per active user beyond airdrop incentives
- 02
Cross-substrate portability (Base → Solana → others)
- 03
Whether DAO-style governance survives or collapses to core team control
Kill shot
What would kill the thesis
If a major Web2 platform (OpenAI, Anthropic, Google) ships proprietary agent rails before crypto-native rails reach institutional adoption, this whole layer becomes a niche.
Coverage
Projects on the radar.
Virtuals Protocol
ConvictionLaunchpad cho on-chain AI agents — biggest agent ecosystem hiện tại trên Base
Deep read →
ai16z / ElizaOS
WatchingAI agent framework open-source — Marc Andreessen's DAO experiment scaling
Deep read →
AIxBT
WatchingSingle-agent crypto Twitter intelligence — first viral 'AI alpha agent'
Deep read →
POV
Long-form in this sector.
Going deeper
Bespoke AI Agents dive for your fund.
Draft thesis · Grey voice in progress, edits land continuously