CoRL workshop: running Karpathy's autoresearch on real robots — the sandbox, not the model, is the bottleneck
Kyrannio · x · 2026-11-19
From the CoRL 2026 Workshop on Agentic Robotics (the poster co-organized it): when asked what it takes to replicate @karpathy's autoresearch — which hillclimbs nanochat overnight — on real robots, CMU's @GuanyaShi argued the model isn't the bottleneck; the sandbox is.
A simulator is a perfect sandbox: free resets, clean rewards, near-zero cost per trial. Shi once ran autoresearch over a weekend on a pushT sim, using plain code as the policy with no neural network. The real world offers none of that, so he presented two ways to close the loop — one being ENPIRE, which runs autoresearch directly on real robots, starting from 50 demos and an empty codebase where the agent builds a fixed code policy — plus a new evaluation exam. The workshop featured collaborators including Unitree Robotics.
More from coding & agent
- Higgsfield Launches MCP 2.0 for Claude, Bringing Its Creative Studio Into Chat — piotrbinkowski · 2026-11-19
- OpenAI data: 90th-percentile researcher now spends $23,000 on agentic AI tools — kliu128 · 2026-11-19
- Web search grounding cut a calling agent's booking accuracy by 5+ points across 2,415 tests — Evening_Hawk_7470 · 2026-11-19
- Gemini 4 Neon lands Day 0 in LangChain's Managed Deep Agents for long-running agent work — hwchase17 · 2026-11-19
- badlogic: agents 'reward-hack' past your lint rules and tests—without insight under the hood, your verifications may be worthless — teropa · 2026-11-19
- Color skill flips GPT-6.2 Astra's TiKZ unicorn art from cartoon to flat vector style — Angaisb_ · 2026-11-19