Illustrative photo about: CryptoSlate: ACE-CaP-0 robot programs succeeded in 221 of 240 trials
Illustrative photo · Photo: jarmoluk / Pixabay, Pixabay Content License
Key figure: CryptoSlate reported that ACE-CaP-0 researchers recorded 221 successes in 240 physical evaluation trials across eight tasks, a success rate of 92.1%.
CryptoSlate reported that ACE-CaP-0 researchers recorded 221 successes in 240 physical evaluation trials across eight tasks, a success rate of 92.1%. Source: cryptoslate.com, 2026-11-15. Source: VERXERA card from cryptoslate.com

What matters

Robot programs completed 221 of 240 physical evaluation trials without a human or coding agent in the execution loop, according to CryptoSlate. The reported result distinguishes execution without that involvement from development, when human guidance and demonstrations were available.

Verified facts

  • CryptoSlate published its report on November 15, 2026, at 2:40 pm GMT.
  • According to CryptoSlate, the ACE-CaP-0 study was posted November 12.
  • CryptoSlate reported that ACE-CaP-0 researchers recorded 221 successes in 240 physical evaluation trials across eight tasks, a success rate of 92.1%.
  • According to CryptoSlate, each task program was frozen—kept unchanged—before 30 separate evaluation trials, with no further updates.
  • CryptoSlate reported that human guidance and demonstrations were available during program development.
  • During evaluation, neither a human nor the coding agent was in the execution loop, according to CryptoSlate.
  • CryptoSlate cautioned that repeated execution on the tested tasks does not establish that the programs can be reused under new conditions.

Analysis

The evaluation described by CryptoSlate concerns repeated execution of eight unchanged task programs, with 30 trials for each. It separates two stages: development with human assistance available, and evaluation without a human or coding agent in the execution loop. The reported 92.1% therefore describes performance across those physical trials, not demonstrated reuse under new conditions.

Editorial opinion

VERXERA opinion: The evidence supports a narrow, meaningful conclusion: repeated execution without a human or coding agent in the loop is distinct from development without human help. In the study posted November 12 and reported by CryptoSlate on November 15, 2026, the programs succeeded in 221 of 240 physical trials, or 92.1%, across eight tasks. Keeping each program unchanged for 30 trials connects that result to repeated execution rather than continuing program updates. Human guidance and demonstrations remained available during development, so the result should not be read as an entirely unassisted process. This account rests on CryptoSlate’s report alone, and reuse under new conditions remains unestablished. Deeper source reading is worthwhile for readers assessing that specific boundary between assisted development and execution.

This is VERXERA’s editorial interpretation, separated from the facts and analysis above. The reader keeps the final judgment.

Counterevidence and limits

CryptoSlate explicitly limits the finding: repeated execution on tested tasks does not establish reuse under new conditions.

Known uncertainties

Whether these task programs can be reused under new conditions remains unresolved in CryptoSlate’s account.