Evidence record
Anthropic R&D Automation Index
August 2026 snapshot · Anthropic · 2026-09-17
Reported result · numeric
No measured subset fully autonomous
Anthropic model R&D work
Work rated fully autonomous · percent
lab reported · extraction review: agent checked
Source and extraction
Published 2026-09-17 · evaluated 2026-08
- Measurements for understanding the pace of AI development inside frontier labsSection 1; Appendix: Measuring AI-led R&DOriginal source ↗
Evaluation setup
| task count | 378 |
|---|---|
| aggregate method | Person-time-weighted automation categories |
| human intervention | Human-directed or supervised work represented by automation levels |
Not reported: split, task snapshot, attempts per task, run count, selection rule, token budget, wall clock budget, hardware, training budget, inference budget, monetary cost, tool access, internet access, filtering, scaffold, evaluator version, task exclusions, contamination concerns, comparability caveats.
Comparability
Not compared with other results.
Limitations
- Self-report using Claude judges. Zero describes fully autonomous measured work, not absence of AI assistance.
- Model judgment and approximate person-time weights.
- Automation share is not a causal productivity estimate.