Method documents
The write-ups behind the numbers, as committed to the repository. Each page carries a short fingerprint of its source file, so a number quoted here can be traced to the exact text.
| Document | File | Last changed |
|---|---|---|
| IDEA-014 D3 and R4 results | docs/methodology/deal-r4-d3-results.md | 2026-11-04 |
| IDEA-014 R4 and D3 specification | docs/methodology/deal-r4-d3-spec.md | 2026-11-04 |
| IDEA-014 R4 v2 results | docs/methodology/deal-r4-v2-results.md | 2026-11-04 |
| IDEA-014 R4 v2 specification | docs/methodology/deal-r4-v2-spec.md | 2026-11-04 |
| IDEA-014 R4 v3 results, and R4 close-out | docs/methodology/deal-r4-v3-results.md | 2026-11-04 |
| IDEA-014 R4 v3 specification | docs/methodology/deal-r4-v3-spec.md | 2026-11-04 |
| deal-v0.1 pre-checks (IDEA-014 M1) | docs/methodology/deal-v0.1-prechecks.md | 2026-11-03 |
| deal-v0.1 results (IDEA-014 M5-M7) | docs/methodology/deal-v0.1-results.md | 2026-11-04 |
| deal-v0.1 specification (IDEA-014 M2) | docs/methodology/deal-v0.1-spec.md | 2026-11-03 |
| Demand pull v2: no detectable gain (IDEA-006, draft, not published) | docs/methodology/demand-pull-v2-result.md | 2026-11-03 |
| Demand pull: no prediction test was run (IDEA-006, draft, not published) | docs/methodology/demand-pull-what-failed-2026-11-03.md | 2026-11-03 |
| Science Lens: extraction accuracy on a fresh sample | docs/methodology/lens-accuracy.md | 2026-11-09 |
| Science Lens: search against keyword baselines | docs/methodology/lens-search-eval.md | 2026-11-09 |
| Patents against v4: result | docs/methodology/patents-v4-results.md | 2026-11-04 |
| Grading a science brief's supported records, in session, with negative controls | docs/methodology/scibrief-grading.md | 2026-11-10 |
| Methodology | docs/methodology.md | 2026-10-27 |
| Data sources (all public) | docs/data-sources.md | 2026-10-26 |
| SX1 results: the science-evaluation comparison, run once | docs/dd/sx1-results.md | 2026-11-10 |
| What I did with 202,638 paper extractions after the model test failed | docs/reports/2026-11-09-extraction-use-report.md | 2026-11-10 |
| A specialist science agent lost to one careful read. Here is what I built instead, and what it measurably does. | docs/reports/2026-11-10-science-eval-report.md | 2026-11-10 |