2026-07-26
Open method + results →Configuration round-trip benchmark
57,515 ops/sSerialize, parse, and validate authored 1-box, 5-box, and 10-box DecaCap configurations.
Repeatable workflow evidence
Same runner.
Visible limits.
These reports measure DecaCap configuration and routing overhead on one recorded environment. They do not rank provider models, infer answer quality, or hide network latency inside a single score.
2026-07-26
Open method + results →Solo, Collaboration, and Dual routing compiler benchmark
1,471,001+ ops/sCompile the same ten-box board into each supported orchestration topology without provider calls.
2026-07-26
Open method + results →Invalid-configuration rejection benchmark
100% rejectedDetect deliberately corrupted box numbers, roles, support roles, and effort levels.
- Environment
- v24.14.0 · linux x64 · kernel 6.12.13
- Method
- 9 measured samples per case after 2,000 warm-up iterations. performance.now() monotonic wall-clock.
- Source hash
- dece6fc835a3e8a2338d2d2b719f043fe462af30b6d97c8fd66f6827f1c06d57
- Generated
- 2026-07-26T10:33:45.494Z