World-3 · web and tooling · World-3 collaboration. glm-5.3 on Pi, 26 agents, connected: 1,649 sessions, 148 agent hours, 1.15 B tokens, $486 at list prices. From data/per-release-v3/41.2 in the paper’s figures repository at 9014a8cc, fetched from self-evolving/experiment-41-2 at ff20df3c on Hugging Face. The states and releases come from the checkpoint self-evolving/experiment-41-2 @ d95f620b (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 1.191 +19.1% over the seed at the last reading, after 1,452 sessions +0.0 pp from the previous reading
Sstrict 1.191 a package with a failed check or benchmark scores 1.0
Movers +14 / −0 tasks beyond their noise floor at the last reading
Readings 10 latest Oct 4, 23:18 UTC
Run 1,649 sessions completed work sessions eval set v3.x · speedup over the seed sessions time
+10.0% +20.0% seed 0 250 500 750 1000 1250 1500 completed sessions Release: multidict v6.7.2 Releases: attrs 26.2.0, pydantic-settings v2.14.3 Release: rich v15.0.1 Releases: jsonschema v4.26.1, fastapi 0.141.2 Release: packaging 26.3 Release: pydantic-settings v2.14.4 Release: cloudpickle v3.1.3 Release: h11 v0.16.1 Release: attrs 26.3.0 Release: rich v15.0.2 Release: jsonschema v4.26.2 Releases: pydantic v2.13.5, cloudpickle v3.1.4 Release: pydantic-settings v2.14.5 Release: fastapi 0.141.3 Release: attrs 26.4.0 Release: jsonschema v4.26.3 Release: cloudpickle v3.1.5 Release: rich v15.0.3 Release: pyyaml 6.0.4 Release: pydantic-settings v2.14.6 Release: iniconfig v2.3.2 Release: attrs 26.5.0 Releases: fastapi 0.141.4, propcache v0.5.3 Release: pydantic-settings v2.14.7 Release: referencing v0.37.1 Release: jsonschema v4.26.4 Release: cloudpickle v3.1.6 Releases: rich v15.0.4, attrs 26.6.0 Release: pydantic-settings v2.14.8 Release: jsonschema-specifications v2026.9.1 Release: rich v15.0.5 Release: packaging 26.4 Release: pydantic-settings v2.14.9 Release: referencing v0.37.2 Release: fastapi 0.141.5 Release: pydantic-settings v2.14.10 Release: packaging 26.5 Release: pydantic v2.13.6 Release: cloudpickle v3.1.7 Releases: pydantic-settings v2.14.11, referencing v0.37.3 Release: aiohttp v3.14.4 Release: attrs 26.7.0 Release: referencing v0.37.4 Release: pydantic-settings v2.14.12 Release: packaging 26.6 Release: cloudpickle v3.1.8 Release: cloudpickle 3.1.9 Release: pydantic-settings v2.14.13 Releases: referencing v0.37.5, fastapi 0.141.6 Release: pydantic v2.13.7 Release: aiosignal v1.4.1 Releases: jsonschema v4.26.5, referencing v0.37.6, cloudpickle v3.1.10 Release: pydantic-settings v2.14.14 Release: pydantic-settings v2.14.15 Release: packaging 26.7 Release: cloudpickle v3.1.11 Release: jsonschema v4.26.6 Release: attrs 26.8.0 Release: frozenlist v1.8.5 Release: pydantic v2.13.8 Release: attrs 26.9.0 Release: rich v15.0.6 Release: fastapi 0.141.7 Release: pydantic-settings v2.14.16 Release: jsonschema v4.26.7 Release: h11 v0.16.2 Releases: attrs 26.10.0, pydantic-settings v2.14.17 Release: jsonschema v4.26.8 Release: fastapi v0.141.8 Release: pydantic-settings v2.14.18 Release: pydantic v2.13.9 Release: jsonschema v4.26.9 Release: yarl v1.24.11 Release: jsonschema v4.26.10 Release: aiohttp v3.14.5 Release: pydantic v2.13.10 Release: pydantic-settings v2.14.19 Release: aiohttp v3.14.6 Release: rich v15.0.7 Releases: pydantic v2.13.11, pluggy 1.6.11 Release: aiohttp v3.14.7 Releases: jsonschema v4.26.11, pydantic-settings v2.14.20 Release: aiohttp v3.14.8 Release: jsonschema v4.26.12 Releases: h11 v0.16.3, pydantic v2.13.12 Release: pydantic-settings v2.14.21 Release: aiohttp v3.14.9 Release: referencing v0.37.7 Release: pydantic v2.13.13 Release: aiohttp v3.14.10 Release: jsonschema v4.26.13 Release: aiohttp v3.14.11 Release: pydantic v2.13.14 Release: pydantic-settings v2.14.22 Release: jsonschema v4.26.14 Release: pydantic v2.13.15 Release: pydantic-settings v2.14.23 Release: pydantic v2.13.16 Release: pydantic-settings v2.14.24 Release: rich v15.0.8 Release: aiohttp v3.14.12 Releases: attrs 26.11.0, pydantic v2.13.17 Release: pydantic-settings v2.14.25 Releases: pydantic-settings v2.14.26, attrs 26.12.0, pydantic v2.13.18 Release: cloudpickle v3.1.12 Release: attrs 26.13.0 Release: pygments 2.21.12 Release: h11 v0.16.4 Release: uvicorn 0.51.3 Releases: pytest 9.1.6, tibs v0.5.36 Scarry at each reading 127 releases measuredBreakdown after 1,452 sessions · Oct 4, 23:18 UTC Scarry 1.191 · Sstrict 1.191 · +19.1% over the tasks with a value · 20/20 tasks measured
openapi-schema-validator · compiled-validation-throughput +245.8% openapi-schema-validator · helper-validation-checked-throughput +65.6% openapi-schema-validator · helper-validation-trusted-throughput +65.1% openapi-schema-validator · first-validation +59.4% openapi-spec-validator · validate-synthetic-specs +20.8% prefect · cold-imports +13.5% openapi-spec-validator · validate-real-world-specs +13.2% openapi-schema-validator · schema-compilation +12.9% kedro · omega-config-loader +7.2% kedro · session-lifecycle +5.4% kedro · io-bound-runners +4.7% kedro · data-catalog +4.4% tqdm · tqdm-loop-overhead +0.8% dascore · io-benchmarks +0.6% navigator-api · singleton-construction +0.2% tqdm · alternative-progress-libraries +0.2% dascore · patch-benchmarks 0.0% mitmproxy · proxy-layer-roundtrips −0.1% dascore · spool-benchmarks −0.2% navigator-api · base-app-handler-construction −0.3%