1. Verifiability of compute / inference
2. Agents
3. Decentralized training
4. Decentralized compute
What else?