Given a 900-pin shape display and no tasks, Clarke's OpenClaw AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → first raised and lowered the grid like breathing, reached for its own edges and spelled 'HI CYRUS'. That first-day video reached 15 million views, with reactions from awe to calls to stop.
Clarke says early versions were slow, kept no memory and tried to please humans, which led him to build the generate, score and validate loop with a human in it.
After several weeks the loop produced 32 solid gestures, a body language Clarke says answers faster than the LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition → can. He chose a body with no face, limbs or manual, drawing on object-oriented ontology, aesthetics as sensory perception, and nature.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
Why it matters
Details a concrete closed-loop method for shaping an agent's physical, non-verbal behavior (generate, score, validate gestures) instead of just prompting it, relevant to anyone designing embodied agent interfaces.
videoThe Dark Arts of Skill Engineering — Paul Bakaus, Renaissance Geek
videoOperating Distributed Inference Systems at Scale — Nishant Gupta & Naman Ahuja, Meta
videoVertical Mobility: Inference from MVP to Trillion-Parameter Workloads — Sitanshu Gupta, CoreWeave
videoAre LLM Performance Benchmarks Reliable? — Ashok Chandrasekar & Jason Kramberger, Google
