Skip to content

WRB scorecard: FF-Occam vs DonSeTch

Captured 2026-09-07 on one Windows machine. Fetch-only. Same WRB pin. Raw JSON stays under gitignored artifacts/wrb/52025c304f6c/results/.

  • FF-Occam: occam.json (2026-09-07T17:25:39), tree 6bc19e5c7c26, long-lived MCP host, OCCAM_WRB_RETAIN_COMPILE=1
  • DonSeTch: donsetch.json (2026-09-07T17:22:36), CLI 3.6.7 (donsetch.exe, git 24e6ee8, npm global)
  • WRB revision: 52025c304f6cdd242eb6d3fef2f0cb3700838fbd
Area Metric FF-Occam DonSeTch Better value
Fetch retrieval 75.0% 87.5% DonSeTch
Fetch tier 1 100.0% 100.0% tie
Fetch tier 2 75.0% 100.0% DonSeTch
Fetch tier 3 38.5% 53.8% DonSeTch
Fetch false positives 0.0% 0.0% tie
Fetch p50 2109 ms 1453 ms DonSeTch
Fetch p90 4313 ms 5918 ms FF-Occam
Fetch median tokens 1250 1028 DonSeTch
Search not measured not measured not comparable
Crawl not measured not measured not comparable

Occam crawl slot remains an occam_map discovery proxy. Search was not run. Do not set OCCAM_SEARCH_PROVIDER=donsetch in a discovery comparison.

WRB uses substring probes and chars / 4 token estimates. This is comparative evidence, not an agent-answer-quality benchmark and not a product-wide success headline.