Outcome benchmarks
No chart until the cohort deserves one.
AI Amigos publishes a benchmark only when at least 10 comparable reports pass privacy and editorial review. Zero reports currently qualify, so no performance claim is shown.
Publication gate
- Same task definition and metric
- Known version and date window
- At least 10 reviewed reports
- No direct or quasi-identifiers
- Median and interquartile range, not a misleading average
- Named reviewer, limitations, and corrections
All cohorts suppressed0 publishable benchmarks
Collection is not evidence. Review, comparability, and the privacy threshold must all pass.
Collection status
| Task | Track | Metric | Reviewed reports | Status |
|---|---|---|---|---|
| AI-assisted support triage | business | median minutes per reviewed case | 0/10 | Suppressed |
| Weekly portfolio proof cycle | careers | rubric score change | 0/10 | Suppressed |
| Assessment with declared AI roles | teaching | rubric alignment score | 0/10 | Suppressed |
| RAG change evaluation | builders | case pass rate | 0/10 | Suppressed |