92.1%: Claude Opus 5.5 Wrote Robot Code That Worked With No AI Watching

A November 12 preprint found robot-control programs written by Claude Opus 5.5 succeeded in 221 of 240 trials, 92.1%, after the code was frozen and the AI left the loop entirely.

Key Takeaways

  • A November 12, 2026 preprint found robot-control programs written by Claude Opus 5.5 succeeded in 221 of 240 trials, a 92.1% success rate, with no AI present during execution.
  • The ACE-CaP-0 study tested eight manipulation tasks on two Agilex PIPERX robot arms using frozen, unedited code across 30 trials per task.
  • Charger insertion scored a perfect 30 of 30, while bread packing was weakest at 25 of 30, likely due to deformable-object variability.
  • Giving the AI access to simulation during development raised success to 100% and cut development time to 2.6 hours, versus 86.7% and 3.7 hours without it.
  • The research points to a hybrid model: use a large AI model once to write deterministic robot code, then run that code cheaply without live AI inference.

92.1%: Claude Opus 5.5 Wrote Robot Code That Worked With No AI Watching

Researchers behind the CaP-X robot-coding benchmark, a group that includes UC Berkeley's Ken Goldberg and Stanford's Fei-Fei Li and Jiajun Wu, posted a preprint on November 12, 2026, showing that programs written by Anthropic's Claude Opus 5.5 succeeded in 221 of 240 physical robot trials, a 92.1% success rate, after the code was frozen with no AI or human present during execution. The study, labeled ACE-CaP-0, ran eight distinct manipulation tasks on a dual-arm robot.

What exactly did the researchers test?

The setup used two Agilex PIPERX robot arms and three Orbbec DaBai DC1 cameras. During a development phase, Claude Opus 5.5 acted as a coding agent: it wrote, ran, inspected and revised its own code based on physical feedback from the robot, with up to 10 human-assisted scene resets allowed per task. Once a program looked solid, the researchers froze it entirely, no more edits, and ran it through 30 fresh evaluation trials per task with neither a person nor the AI intervening. The frozen code had to react to live camera and sensor data on its own.

That's a meaningfully different test than most robot-AI demonstrations, which typically keep a neural network running continuously during execution. Here, the AI's job ended once development stopped.

Which tasks were hardest, and why?

Charger insertion came out perfect at 30 of 30 trials. Bread packing was the weakest performer at 25 of 30. In between, bottle uncapping, flower insertion and shoe packing each hit 27 of 30, cup nesting and parcel reorientation landed at 28 of 30, and chicken plating reached 29 of 30. The pattern suggests that tasks involving deformable or inconsistently shaped objects, bread, flowers, uncapped bottles, are harder for a fixed set of decision rules to handle reliably than tasks with rigid, predictable geometry like a charger and its port.

Why does freezing the code matter?

Most current approaches to robot manipulation rely on vision-language-action models that run inference continuously, frame by frame, during execution. That's computationally expensive and requires the model to be present and running every time the robot acts. The Code-as-Policies approach tested here flips that: a capable AI model is used once, intensively, during development, then steps aside. Execution afterward is just a program running, cheap and deterministic.

The researchers also found that giving the coding agent access to simulation during development mattered a lot. Trials developed with simulation hit 100% success and took an average of 2.6 hours to develop, versus 86.7% success and 3.7 hours without it. Simulation access, in other words, made Claude Opus 5.5 both faster and more reliable at writing working robot code.

What does this mean for the robotics industry's foundation-model push?

Companies like Figure, Physical Intelligence and Nvidia, with its GR00T platform, are pouring resources into end-to-end models meant to generalize across open-ended tasks without task-specific programming. The ACE-CaP-0 results suggest a complementary path for well-defined, repetitive jobs: let a large coding model write a deterministic program once, verify it in simulation, then deploy that program without needing a model running at inference time. That could cut runtime compute costs sharply for the narrow, repeatable tasks that make up much of warehouse and light-manufacturing work, even as general-purpose VLA models remain the better bet for tasks nobody has scripted before.

Category: Machine Learning

Tags: robot manipulation Claude Opus 5.5 Anthropic Code-as-Policies Ken Goldberg Agilex PIPERX

Related Articles

Frequently Asked Questions

What is ACE-CaP-0?

A November 2026 study by robotics researchers testing whether code written by Anthropic's Claude Opus 5.5 could control a robot reliably after being frozen, with no AI or human intervention during execution.

What was the overall success rate?

Claude Opus 5.5-written robot programs succeeded in 221 of 240 physical trials across eight manipulation tasks, a 92.1% success rate.

Which task had the lowest success rate?

Bread packing, at 25 of 30 trials, the weakest result among the eight tasks tested, likely due to the deformable, inconsistent shape of bread compared to rigid objects.

Why does it matter that the code was frozen during testing?

It shows the AI's involvement can end after development, with the resulting program running on its own, unlike vision-language-action models that typically require continuous AI inference during robot operation.

Who conducted the research?

A team including UC Berkeley's Ken Goldberg and Stanford's Fei-Fei Li and Jiajun Wu, associated with the CaP-X robot-coding benchmark project.