Patterns
| Pattern | What it means | Why it helps |
|---|---|---|
| Champion/challenger | Compare a candidate with the current best artifact. | Keeps improvement grounded. |
| Specialist population | Maintain narrow high-value models. | Reduces waste and improves fit. |
| Local-first inference | Run capable tasks on controlled hardware. | Preserves private work and reduces latency. |
| Evidence packet | Bundle scorecard, lineage, and release notes. | Makes adoption explainable. |
| Lineage DAG | Parent-child graph for artifacts. | Turns experiments into memory. |
| Resource ledger | Tracks latency, memory, energy, and cost. | Keeps the ecology frugal. |
| Human review loop | Names where people shape outcomes. | Converts expertise into durable capability. |
| No-op decision | Keep current state when change does not add value. | Preserves quality and focus. |
| Reversible release | Use shadow, canary, limited release, and rollback. | Encourages experimentation with evidence. |
| Mutualist score | Measure whether the ecology strengthens people. | Keeps benefits at the center. |
Source reports used for this guide
These reports are preserved verbatim in the site archive. The guide above is an editorial synthesis and may narrow, qualify, or reorganize claims from the source material.