The Model Race Hangover: Why the AI Loser May Win the Infrastructure War, and How to Build Your Own Exit
The AI labs spent three years racing to build the best model. That race is largely over, open-weight models now match proprietary ones on most benchmarks, and they are becoming useful for almost every use case. What matters next is not who has the best intelligence, but who controls the context around it and who owns the compute delivering it.
Co-founder & CTO at Finisterra Labs
Alfonso de la Rocha is the Co-founder and CTO of Finisterra Labs, where he leads the development of infrastructure for organizing and querying the world’s structured data, including the company’s Baselight database. His career centers on distributed systems and blockchain technology, having spent over four years at Protocol Labs working on some of the most demanding problems in the Web3 ecosystem. There he was a core contributor to InterPlanetary Consensus (IPC), a framework for scaling blockchain networks through hierarchical subnets, as well as to Filecoin’s indexing and content-routing systems and the Go-f3 fast finality protocol.
His areas of expertise span consensus algorithms, decentralized infrastructure, advanced cryptography and the broader Filecoin and IPFS stack, and he writes regularly about distributed systems, AI and decentralization on his personal blog under the handle adlrocha. Through his research and engineering work he has helped move ambitious ideas in consensus and scalability from prototypes to live testnet deployments, and he continues to build at the intersection of distributed systems, data infrastructure and emerging technology.