Nvidia CEO Jensen Huang declared AGI is already here, then partially walked it back — redefining the term to fit current AI agent capabilities.
On the Lex Fridman podcast, Nvidia CEO Jensen Huang stated 'I think we've achieved AGI' when asked when AGI would arrive. Fridman defined AGI as an AI system capable of starting, growing, and running a $1B+ tech company. Huang cited OpenClaw, an open-source AI agent platform with viral adoption, as evidence. He then walked back the claim, noting that '100,000 of those agents building Nvidia is zero percent' likely — suggesting current AI still falls far short of true AGI by most definitions.
Huang's 'AGI' claim is definitional sleight-of-hand, not a technical announcement. No new APIs, models, or capabilities were released. The actual signal buried here is that AI agent platforms like OpenClaw are hitting viral adoption numbers — which does matter for what you're building on top of agent frameworks today.
Pull OpenClaw's GitHub repo this week and benchmark its agent task-completion rate against LangChain or CrewAI on a workflow you already have in production — if it hits >80% on your evals, it's worth a swap.
Go to github.com and search 'OpenClaw AI agent'
Check the stars, recent commits, and open issues tab to assess community health
Compare the README's benchmark claims against your current agent framework's documented performance
A side-by-side note on OpenClaw vs your current stack: community size, last commit date, and one concrete capability difference
Tags