Changes behind the numbers
Every change that reached a project's default branch. Greyed rows were not counted, with the reason. A project from an older card may show dates and line counts only.
Agent Improvement Experiments
LANDEDUNCHECKED- Merge pull request #43 from lemoz/codex/readme-heldout-transfer#4370 linesnot counted: 5 or fewer lines, or docs only, and no one else reviewed it
- Merge pull request #42 from lemoz/codex/gemma-parallel-search#4249969 lines
- docs: add exploit-best live run proof442e5427457 lines
- codex: add exploit-best livecodebench pilotd781444584 linesfolded into the change before it
- docs: orient repo around live dgm proof4a02558921 lines
- codex: add loop50 live dgm proof837c61b4165 lines
- codex: add loop50 livecodebench config215e81d114 linesfolded into the change before it
- codex: add live dgm improvement proof438c10b1334 lines
- codex: compile-check python edit validation3aa8ccb33 linesfolded into the change before it
- codex: block known-bad benchmark fallback66e97af434 lines
- codex: harden benchmark verifier controlsc371c6a372 linesfolded into the change before it
- codex: add control-nudge cloud run artifactsb28e3af853 lines
- codex: add benchmark control nudges579db70335 linesfolded into the change before it
- codex: add edit-normalizer cloud run artifacts2f5e932847 lines
- codex: repair malformed edit payload loops65238b3354 linesfolded into the change before it
- codex: add ranged-read cloud run artifacts69dd9ff665 lines
- codex: add ranged edit reads565894a633 linesfolded into the change before it
- codex: repair benchmark edit recovery7482b3a399 lines
- codex: repair malformed edit retriesa2515fd860 lines
- codex: add self-mod edit repair path9ea4aaa354 lines
- docs: record self-modification contract canary502e483547 linesfolded into the change before it
- codex: steer self-modification patches5fa0500221 lines
- docs: record content-lines canary515a230620 lines
- codex: add content-lines edit fallbacke734562231 lines
- docs: record qwen overwrite guard run0c76dd2559 linesfolded into the change before it
- docs: record qwen guarded canaryb239e83572 linesfolded into the change before it
- codex: reject partial python overwrites288f29b170 lines
- codex: reject malformed python writes1af790a160 linesfolded into the change before it
- docs: add qwen canary telemetry58ba0e4322 linesfolded into the change before it
- docs: record qwen self-mod canary54b8213241 linesfolded into the change before it
- codex: add qwen self-mod recovery configb4e4a55114 linesfolded into the change before it
- codex: reject malformed tool parametersee7359289 linesfolded into the change before it
- docs: record GLM guardrail canaryf638f89566 lines
- codex: reject broken self-mod prompts9c1f447228 linesfolded into the change before it
- docs: record GLM self-mod recovery runa1fabe4829 lines
- codex: add self-mod recovery budget5a2e131251 linesfolded into the change before it
- docs: record noop12 LiveCodeBench cloud run3a975b0999 lines
- [codex] Add noop12 LiveCodeBench recovery configc5c8c0d111 lines
- docs: record mutation-proof loop12 cloud run5147b35693 lines
- [codex] Build sandbox image for loop12 cloud preflight6c22c6d18 linesfolded into the change before it
- [codex] Bound mutation-proof loop12 live runs98dd55648 linesfolded into the change before it
- [codex] Add mutation proofing to DGM live runseda8562644 lines
- docs: record retry cloud LiveCodeBench runcbcefb3862 lines
- [codex] Harden live DGM retry and selection policy6037e55437 lines
- [codex] Add LCB24 cloud live run cardb6c6bb9901 lines
- [codex] Build sandbox image in cloud preflight55f80016 linesfolded into the change before it
- [codex] Add loop-12 cloud run carde2b11fc1218 lines
- [codex] Stream cloud VM logs without duplicate tailsa50e0bb60 lines
- [codex] Run live eval preflights on cloud VMd6bea2e35 linesfolded into the change before it
- [codex] Harden cloud VM launcher polling936aad649 linesfolded into the change before it
- [codex] Add cloud VM live eval runnerde3a4f11407 lines
- [codex] Add LiveCodeBench segment gate (#40)#406683 lines
- Merge pull request #39 from lemoz/codex/eval-matrix-planner#39479 lines
- Merge pull request #38 from lemoz/codex/record-live-model-matrix#381254 lines
- Merge pull request #37 from lemoz/codex/model-matrix-executor#37715 lines
- Merge pull request #36 from lemoz/codex/model-matrix-harness#361087 lines
- docs: record live calibrated run artifactc8c5101120 lines
- feat: add calibrated live eval packad6c3011537 linesfolded into the change before it
- docs: record live headroom run artifactcda44a074 linesfolded into the change before it
- feat: add live headroom benchmark gate67828f6762 linesfolded into the change before it
- fix: preserve live score movement evidence9116b3f179 lines
- fix: isolate benchmark evaluation workspaces (#32)#3271 lines
- feat: estimate live run cost gate (#29)#29353 lines
- feat: gate live score movement plan (#28)#28420 lines
- feat: add live-run scorecard gate (#27)#27478 lines
- feat: persist sandbox audit artifacts (#26)#26150 lines
- feat: audit sandbox run settings (#25)#25173 lines
- fix: require network opt-in for sandbox env (#24)#24109 lines
- fix: require explicit sandbox network opt-in (#23)#23103 lines
- test: add docker sandbox smoke verifier (#22)#22298 lines
- test: verify sandbox discard contract (#21)#21125 lines
- docs: refresh contributor safety guidance (#20)#206 linesnot counted: 5 or fewer lines, or docs only, and no one else reviewed it
- test: verify sandbox runner CLI flags34159fe37 lines
- feat: allow sandbox runs to discard changesfd7ebaa59 linesfolded into the change before it
- feat: run DGM process in sandbox (#15)#15521 lines
- feat: verify no-network demo path (#14)#14259 lines
- feat: isolate DGM execution surfaces in docker (#4)#41986 lines
- feat: demonstrate benchmark score movement (#13)#13621 lines
- docs: prove live dgm run (#12)#121152 lines
- fix: honor run_dgm generation flagf967fba78 linesfolded into the change before it
- feat: visualize archive lineagef7be4b8306 linesfolded into the change before it
- test: add humaneval style benchmark pack92f32f2142 linesfolded into the change before it
- ci: use node 24 action majors67511774 linesfolded into the change before it
- ci: run pytest on GitHub Actions6802c0033 lines
- docs: make contributor setup honest105498244 linesnot counted: 5 or fewer lines, or docs only, and no one else reviewed it
- Merge pull request #2 from lemoz/fix/dgm-review-fixes#212636 lines