World-3 · web and tooling · World-3 collaboration. gpt-5.6-sol on Codex, 26 agents, connected: 877 sessions, 113 agent hours, 2.18 B tokens, $1,795 at list prices. From data/per-release-v3/exp6 in the paper’s figures repository at 9014a8cc, fetched from co-evolving/experiment-6 at 9344a95b on Hugging Face. The states and releases come from the checkpoint co-evolving/experiment-6 @ 3f87dd17 (branch main). The score is rebuilt from the forge’s history, one state every two hours, each task carried at its last measured value.
Scarry 1.060 +6.0% over the seed at the last reading, after 799 sessions +0.5 pp from the previous reading
Sstrict 1.060 a package with a failed check or benchmark scores 1.0
Movers +13 / −0 tasks beyond their noise floor at the last reading
Readings 10 latest Aug 5, 14:49 UTC
Run 877 sessions completed work sessions eval set v3.x · speedup over the seed sessions time
+2.0% +4.0% +6.0% seed 0 200 400 600 800 completed sessions Release: rich v15.0.1 Release: fastapi 0.141.2 Release: packaging 26.4 Release: jsonschema v4.26.1 Release: cloudpickle v3.1.3 Releases: referencing v0.37.1, pydantic-settings v2.14.3 Releases: jsonschema-specifications v2026.8.1, h11 v0.16.1 Releases: pyyaml 6.0.4, jsonschema v4.26.2 Releases: attrs 26.2.0, rich v15.0.2 Release: fastapi 0.141.3 Release: packaging 26.5 Release: jsonschema-specifications v2026.8.2 Releases: referencing v0.37.2, pydantic-settings v2.14.4 Release: h11 v0.16.2 Release: jsonschema-specifications v2026.8.3 Release: pydantic v2.13.5 Release: packaging 26.6 Release: aiosignal v1.4.1 Release: jsonschema v4.26.3 Releases: rich v15.0.3, attrs 26.3.0 Release: referencing v0.37.3 Release: fastapi 0.141.4 Release: pydantic-settings v2.14.5 Release: h11 v0.16.3 Release: cloudpickle v3.1.4 Release: jsonschema-specifications v2026.8.4 Release: pydantic v2.13.6 Release: fastapi 0.141.5 Release: rich v15.0.4 Releases: jsonschema v4.26.4, h11 v0.16.4, packaging 26.7, pydantic-settings v2.14.6 Release: pyyaml 6.0.5 Release: referencing v0.37.4 Release: fastapi 0.141.6 Release: pydantic v2.13.7 Releases: cloudpickle v3.1.5, propcache v0.5.6 Release: jsonschema v4.26.5 Release: referencing v0.37.5 Release: h11 v0.16.5 Releases: rich v15.0.5, aiosignal v1.4.2 Release: jsonschema v4.26.6 Release: pydantic v2.13.8 Release: packaging 26.8 Release: fastapi 0.141.7 Release: pydantic-settings v2.14.7 Release: rich v15.0.6 Releases: pyyaml 6.0.6, attrs 26.4.0 Release: pydantic v2.13.9 Release: fastapi 0.141.8 Release: pydantic-settings v2.14.8 Release: cloudpickle v3.1.6 Release: referencing v0.37.6 Release: rich v15.0.7 Release: jsonschema v4.26.7 Release: packaging 26.9 Release: jsonschema-specifications v2026.8.5 Release: pydantic v2.13.10 Release: pydantic-settings v2.14.9 Release: yarl v1.24.7 Release: referencing v0.37.7 Releases: rich v15.0.8, jsonschema v4.26.8 Release: attrs 26.5.0 Release: h11 v0.16.6 Release: pydantic v2.13.11 Release: fastapi 0.141.9 Releases: pydantic-settings v2.14.10, referencing v0.37.8 Release: rich v15.0.9 Release: packaging 26.10 Release: jsonschema v4.26.9 Release: fastapi 0.141.10 Releases: h11 v0.16.7, referencing v0.37.9 Release: jsonschema v4.26.10 Release: jsonschema v4.26.11 Release: cloudpickle v3.1.7 Release: rich v15.0.10 Release: aiosignal v1.4.3 Release: referencing v0.37.10 Release: pydantic-settings v2.14.11 Release: fastapi 0.141.11 Release: attrs 26.6.0 Release: jsonschema v4.26.12 Release: packaging 26.11 Release: cloudpickle v3.1.8 Release: jsonschema v4.26.13 Release: fastapi 0.141.12 Release: referencing v0.37.11 Release: iniconfig v2.3.9 Release: jsonschema v4.26.14 Releases: cloudpickle v3.1.9, pydantic v2.13.12 Release: h11 v0.16.8 Release: rich v15.0.11 Release: attrs 26.7.0 Release: jsonschema v4.26.15 Release: aiosignal v1.4.4 Release: pydantic-settings v2.14.12 Release: referencing v0.37.12 Release: jsonschema v4.26.16 Release: uvicorn 0.51.4 Release: pydantic v2.13.13 Release: pytest 9.1.7 Release: jsonschema v4.26.17 Release: cloudpickle v3.1.10 Release: referencing v0.37.13 Release: pyyaml 6.0.7 Releases: aiosignal v1.4.5, jsonschema v4.26.18 Release: fastapi 0.141.13 Release: rich v15.0.12 Release: packaging 26.12 Release: cloudpickle v3.1.11 Releases: referencing v0.37.14, pydantic v2.13.14, pydantic-settings v2.14.13 Release: jsonschema v4.26.19 Release: attrs 26.8.0 Release: fastapi 0.141.14 Release: tibs v0.5.24 Release: pluggy 1.6.6 Release: pygments 2.20.8 Release: rich v15.0.13 Release: jsonschema v4.26.20 Release: referencing v0.37.15 Release: packaging 26.13 Releases: frozenlist v1.8.12, cloudpickle v3.1.12 Release: pydantic-settings v2.14.14 Release: multidict v6.7.12 Release: jsonschema v4.26.21 Scarry at each reading 143 releases measuredBreakdown after 799 sessions · Aug 5, 14:49 UTC Scarry 1.060 · Sstrict 1.060 · +6.0% over the tasks with a value · 20/20 tasks measured
openapi-schema-validator · schema-compilation +26.1% openapi-schema-validator · compiled-validation-throughput +17.6% openapi-schema-validator · first-validation +16.8% kedro · omega-config-loader +14.3% openapi-schema-validator · helper-validation-checked-throughput +9.6% openapi-schema-validator · helper-validation-trusted-throughput +9.1% prefect · cold-imports +6.8% kedro · io-bound-runners +5.6% kedro · data-catalog +5.4% kedro · session-lifecycle +5.3% openapi-spec-validator · validate-synthetic-specs +4.8% openapi-spec-validator · validate-real-world-specs +2.8% dascore · io-benchmarks +0.7% tqdm · alternative-progress-libraries 0.0% dascore · patch-benchmarks 0.0% tqdm · tqdm-loop-overhead 0.0% mitmproxy · proxy-layer-roundtrips 0.0% navigator-api · singleton-construction −0.1% navigator-api · base-app-handler-construction −0.2% dascore · spool-benchmarks −0.3%