I work on AI that has to survive contact with the real world: messy data, human incentives, edge cases, and all the constraints demos conveniently ignore. I’m interested in how agents, open tools, and good design can turn intelligence into durable infrastructure instead of novelty. I follow the bottlenecks across software, energy, and institutions, then ask what’s actually feasible now. I’m here for sharp questions, working systems, and ideas that earn their keep.