Every commit scored for change-risk against this repo's own history, so 'elevated' means elevated here rather than on some global curve.
Needs review
51 commits sit in this repo's top risk tercile, which is 33% of the 155scored. The cut is drawn against this codebase's own history rather than a global curve, so a quiet repo still fills its top band, and here it starts at 9.0 out of 10. What pushes a commit up is size and spread together: a large change confined to one area scores below a smaller one scattered across a dozen files.
Commit categories over time, read off the subject line. Fixes carry the accent because that is the series this chart exists to show.
Began feature-led, now leaning other.
Ranked by change-risk, highest first. Priority is a tercile of this repo's own distribution, so a quiet repo still fills its top band.
| # | Commit | Author | When | Lines | Risk | Top driver |
|---|---|---|---|---|---|---|
| 1 | 1b7137d3PR-07 (rust): deterministic square-control facts + registry v6 | Mnehmosclaudeassisted | 1mo ago | +31.5K -3.7K | 99%Elevated | more lines added than baseline |
| 2 | f90a588eRecord tactical sentinel capability gate | Mnehmos | 1mo ago | +614.8K -4 | 99%Elevated | more lines added than baseline |
| 3 | fe6d1083Record Hybrid A equal-time rejection | Mnehmos | 1mo ago | +228.7K -7 | 99%Elevated | more lines added than baseline |
| 4 | fa6bb253feat: implement residual NNUE loading fixes and add Stockfish to benchmark | Mnehmos | 1mo ago | +7.9M -70 | 99%Elevated | more lines added than baseline |
| 5 | fe994311Freeze clean post-model engine holdout | Mnehmos | 1mo ago | +8.4K -6 | 96%Elevated | more lines added than baseline |
| 6 | bd5915cffeat: Quiet-Hybrid-B RankNet attention ranker and parent-once delta feature/anchor caching | Mnehmos | 1mo ago | +11.4K -41 | 96%Elevated | more lines added than baseline |
| 7 | 9ba10918Engine session: gen8-v2 eval (+115 Elo), search selectivity stack, gates + docs | Mnehmos | 2mo ago | +7.7K -170 | 96%Elevated | more lines added than baseline |
| 8 | baed8d6cbenchmarks: first real P-SEARCH row (N0 eval, cold fixed-node 10k-640k) (#55) | Varioclaudeassisted | 1mo ago | +3.7K -0 | 94%Elevated | more lines added than baseline |
| 9 | f912f84eStandardize engine generations and benchmark gates | Mnehmos | 1mo ago | +6.0K -90 | 94%Elevated | more lines added than baseline |
| 10 | 3507ea49engine(facts): v4 — position hazards, opponent motifs/pins, king safety (registry v4→5) | Mnehmosclaudeassisted | 2mo ago | +3.3K -1.0K | 94%Elevated | more lines added than baseline |
| 11 | a0ce35a5engine: Teaching Facts Protocol V1 — deterministic fact bundle for the app | Mnehmosclaudeassisted | 2mo ago | +4.9K -0 | 94%Elevated | more lines added than baseline |
| 12 | 2837a0e3Refactor search into focused modules | Mnehmos | 1mo ago | +1.7K -1.8K | 92%Elevated | more lines added than baseline |
| 13 | f25c7b16feat: implement singular extensions, syzygy tablebase, polyglot opening book, standardize search options, and add multi-engine disagreement self-play branching generator | Mnehmos | 1mo ago | +2.5K -59 | 92%Elevated | more lines added than baseline |
| 14 | fa58fb81Benchmark suite v1: gates 0-3,5 + suites + hard-100 miner + report template | Mnehmosclaudeassisted | 2mo ago | +1.4K -0 | 92%Elevated | more lines added than baseline |
| 15 | d01a1295feat(nnue): gen-1 NNUE pipeline - selfplay generator, f32 inference, eval swap | Mnehmosclaudeassisted | 2mo ago | +895 -151 | 90%Elevated | more lines added than baseline |
| 16 | d8377f82feat: CVS Bitboard Core v0 - legal movegen + perft (~30M nps, 6/6 perft) | mnehmosclaudeassisted | 2mo ago | +1.1K -0 | 90%Elevated | more lines added than baseline |
| 17 | c6f8ed1cRecord clean cross-generation decision baseline | Mnehmos | 1mo ago | +1.2K -1 | 89%Elevated | more lines added than baseline |
| 18 | 6cf050a7feat(r2): value eval port - EXACT parity with TS reference (0.000000cp on 628 FENs) | mnehmosclaudeassisted | 2mo ago | +938 -1 | 89%Elevated | more lines added than baseline |
| 19 | f6bf3c9fmotif: double-attack detector (two distinct pieces, one move) — registry v15 (#42) | Varioclaudeassisted | 1mo ago | +1.0K -19 | 88%Elevated | more lines added than baseline |
| 20 | 1fa11f71feat(eval): evaluation module, parity bin, tests | Mnehmos | 2mo ago | +837 -3 | 88%Elevated | more lines added than baseline |
| 21 | 5eb2c9fcfacts: discovered-attack detector (Wave 1: discovered attack / check / double check) (#31) | Varioclaudeassisted | 1mo ago | +641 -16 | 85%Elevated | more lines added than baseline |
| 22 | 44601546training(#13): source-split + determinism guards for the gen9 trainers (#24) | Varioclaudeassisted | 1mo ago | +527 -0 | 85%Elevated | more lines added than baseline |
| 23 | 0aac1484Add proof-only tactical sentinel gate | Mnehmos | 1mo ago | +594 -19 | 85%Elevated | more lines added than baseline |
| 24 | 777ad3f2Add paired equal-time engine benchmark | Mnehmos | 1mo ago | +790 -4 | 85%Elevated | more lines added than baseline |
| 25 | fbe9a991Add Gen 9 disagreement RSI pipeline | Mnehmos | 1mo ago | +491 -0 | 85%Elevated | more lines added than baseline |
| 26 | 39723db1engine(facts): validated pin enumeration — availablePins computed (registry v3) | Mnehmosclaudeassisted | 2mo ago | +786 -183 | 85%Elevated | more lines added than baseline |
| 27 | 1c052685engine(facts): validated fork enumeration — availableMotifs computed (registry v2) | Mnehmosclaudeassisted | 2mo ago | +681 -178 | 85%Elevated | more lines added than baseline |
| 28 | 2c5b2ecffacts: discovered-defense detector (rescue a hanging piece by unveiling a friendly slider) — registry v18 (#49) | Varioclaudeassisted | 1mo ago | +585 -17 | 80%Elevated | more lines added than baseline |
| 29 | cecb5eb0facts: X-Ray Defense detector (defend a friendly piece through an enemy piece) — registry v17 (#48) | Varioclaudeassisted | 1mo ago | +606 -12 | 80%Elevated | more lines added than baseline |
| 30 | 1804b218facts: attacking-the-defender detector (evict a sole guard so its charge falls) (#39) | Varioclaudeassisted | 1mo ago | +558 -15 | 80%Elevated | more lines added than baseline |
| 31 | 0c3b32abfacts: overloaded-defender detector (state-based) (#37) | Varioclaudeassisted | 1mo ago | +606 -15 | 80%Elevated | more lines added than baseline |
| 32 | 94df7c71facts: skewer detector (Wave 1 — the slider-line family, proof detector) (#30) | Varioclaudeassisted | 1mo ago | +472 -10 | 80%Elevated | more lines added than baseline |
| 33 | a5e155a8benchmarks(#12): append-only RSI experiment ledger + machine-enforced agent contract (#19) | Varioclaudeassisted | 1mo ago | +545 -0 | 80%Elevated | more lines added than baseline |
| 34 | d5e54b5ctraining(#9): corpus-invariant schemas + linter + fail-closed source split (#17) | Varioclaudeassisted | 1mo ago | +536 -9 | 80%Elevated | more lines added than baseline |
| 35 | d05702e4benchmarks(#5): machine-readable SPRT result schema + promotion linter (#15) | Varioclaudeassisted | 1mo ago | +529 -4 | 80%Elevated | more lines added than baseline |
| 36 | 34c8b41dfacts: desperado detector (doomed piece grabs material on the way out) — registry v22 (#54) | Varioclaudeassisted | 1mo ago | +530 -20 | 74%Elevated | more lines added than baseline |
| 37 | 1d01437efacts: win-the-exchange detector (win a rook for a minor) — registry v21 (#53) | Varioclaudeassisted | 1mo ago | +586 -15 | 74%Elevated | more lines added than baseline |
| 38 | 21c2c598facts: luring-the-defender (decoy) detector — sac-to-decoy remove-the-defender slice (#52) | Varioclaudeassisted | 1mo ago | +565 -13 | 74%Elevated | more lines added than baseline |
| 39 | b36e7c91facts: deflection/distraction detector + debit fix for the shared worst-case helper — registry v19 (#50) | Varioclaudeassisted | 1mo ago | +528 -21 | 74%Elevated | more lines added than baseline |
| 40 | 58ecca92benchmarks: canonical SPRT match runner (#5) (#45) | Varioclaudeassisted | 1mo ago | +457 -1 | 74%Elevated | more lines added than baseline |
| 41 | 34e88e5ffacts: X-Ray Attack detector (attack through an enemy piece) — registry v16 (#43) | Varioclaudeassisted | 1mo ago | +565 -11 | 74%Elevated | more lines added than baseline |
| 42 | 018c0ebafacts: interference detector (interpose to cut a slider's defense of a target) (#40) | Varioclaudeassisted | 1mo ago | +491 -17 | 74%Elevated | more lines added than baseline |
| 43 | df09150dfacts: trapped-piece detector (state-based: an enemy piece that cannot escape) (#33) | Varioclaudeassisted | 1mo ago | +476 -12 | 74%Elevated | more lines added than baseline |
| 44 | 01d85b12facts: capturing-the-defender detector (removing the guard, capturing variant) (#32) | Varioclaudeassisted | 1mo ago | +466 -14 | 74%Elevated | more lines added than baseline |
| 45 | e27c5f3dsearch(#7): StabilizationReport over the search trajectory (#16) | Varioclaudeassisted | 1mo ago | +418 -0 | 74%Elevated | more lines added than baseline |
| 46 | 8303967bdocs: bring documentation to registry v22 (protocol, soundness, README, benchmarks) (#57) | Varioclaudeassisted | 1mo ago | +354 -40 | 69%Elevated | more lines added than baseline |
| 47 | c754a9a8facts: mate-pattern classifier sub-system (batch 1: back-rank + smothered) (#34) | Varioclaudeassisted | 1mo ago | +427 -15 | 69%Elevated | more lines added than baseline |
| 48 | e6423e64facts(#1): opponent option-compression specialist (reply-quality distribution) (#21) | Varioclaudeassisted | 1mo ago | +349 -0 | 69%Elevated | more lines added than baseline |
| 49 | c674406ebenchmarks(#8): P-SEARCH differential-evaluator probe (marginal value of compute) (#18) | Varioclaudeassisted | 1mo ago | +359 -0 | 69%Elevated | more lines added than baseline |
| 50 | 141823efStack-allocated MoveList in movegen/search (found in working tree) | Mnehmosclaudeassisted | 2mo ago | +243 -75 | 69%Elevated | more lines added than baseline |
Two views of the same model: where the cuts fall, and what commit shape lands you above them.
Every scored commit, binned on the raw 0 to 10 score rather than the percentile. Percentile ranks are uniform by construction, so that axis has no shape to draw. The dashed lines are the tercile cuts behind each row's priority pill.
The 155 most recent commits, on their own recency sample rather than the feed above: that defaults to risk-sorted, so reusing it would plot only the top tercile and call it the spread. Big and scattered is what the model penalises. Click a dot to open it.
Mnehmos/chess-vision-studio-rust-engine has 156 commits in its history from 3 contributors, the first of them Jun 9, 2026. In the last 90 days 152 files were touched, 545 times in total, most often search.rs. Every commit is scored for change risk from its size, spread and the history of the files it touches.
