Roadmap And Contributions
Near term legibility, mid term operator composition, long term measured autonomy with open tasks.
Near term focus
Near term work favors legibility and small safe wins. The diary format gains richer cost views. The fixture set grows across math, code repair, and planning tasks. Judge calibration gains a visible dashboard.
Docs gain guided paths for students, builders, and reviewers. Each path keeps plain words, steady rhythm, and calm color. New diagrams favor CSS and SVG so pages stay fast.
Mid term composition
Mid term work composes data, harness, and model loops under joint budgets. Transfer tests cover new domains with true held out cases. The archive gains lineage views that show which old win enabled a later breakthrough.
Meta scheduling revises operator order from evidence rather than habit. Cost aware routing sends cheap tries to fast loops and costly tries to gated loops. Review load stays human scale through small diffs.
Long term measure
Long term work admits broader tasks only with stronger sandboxes and clearer stop rules. Governance grade measurement tracks autonomy, grounding strength, and drift with the same rigor as accuracy.
Contributions stay welcome through small focused pull requests with tests attached. Open an issue first for larger ideas. Keep judges separate from workers. Keep costs visible. Keep language simple so the record stays shared.
Open tasks by horizon
| Horizon | Task | Proof of done |
|---|---|---|
| Near | Cost dashboard in diary | Before and after cost per generation |
| Near | Guided student path | Three readable pages with diagrams |
| Mid | Joint operator schedule | Shared budget with logged choices |
| Mid | Lineage view | Archive graph from win to later win |
| Long | Autonomy metrics | Grounded scorecard with drift checks |