We’re a small, research-driven, VC-backed team building NeoHuman: a desktop-first system that observes how people actually use AI tools, files, and workflows during real tasks, and turns those sessions into reliable evaluation evidence. We care about the process, not just the final answer — what someone typed, what they copied, what they verified, how they moved across apps. We’ve published the AgentIF-OneDay benchmark. Now we’re scaling the capture and evaluation infrastructure underneath it.