Anthropic Cookbook Engineering Benchmarks
Anthropic Cookbook delivery benchmarks are based on 5 merged pull requests in the last 30 days (114 all-time), with 0.5 hours median cycle time, 0.1 hours median review time, and 3 active contributors. AI-related signals appear on 40.0% of recent work. PR volume is down -38% compared to the prior period.
Repo vs Anthropic Average
Delivery Trend and AI Signals
PR Backlog (Humans)
Excludes AI agents and automation accounts.
PR Backlog (AI Agents)
AI coding agents only.
AI Signals vs Cycle Time
Correlation
Cycle Time by AI-Related Signals
PR Breakdown
Top AI Tools
PR Type Breakdown
Recent Pull Requests
View PR Explorer| Title | Author | AI |
|---|---|---|
| Bump cookbook defaults to Opus 5.5, Sonnet 5.5 and Fable 5.1 | cj-ant | AI |
| feat(managed_agents): add agent security review cookbook | cj-ant | — |
| feat(observability): add OpenTelemetry tracing cookbook | cj-ant | — |
| feat(managed_agents): move plan-big-execute-small to multia… | nikblanchet | AI |
| docs(cost_optimization): add model-agnostic caveats for pre… | benlehrburger-ant | — |
Methodology
Metrics are computed from merged pull request data synced daily from GitHub. AI-related metrics are directional signals inferred from pull request activity patterns and heuristics. Summary metrics use a 30-day rolling window; trends use a 90-day window. Last updated 20 hours, 39 minutes ago.
Updated daily from public GitHub pull requests.
See Your Team's Benchmarks