Two runs, side by side

20 chore votes are left out of the change: missing or skipped in one run, or scored on other inputs.

Model	Base	Compared	Change	Chores changed
Claude Opus 5.5	20.0	20.0	no change over 10 chores	none
Claude Sonnet 5.5	5.0	5.0	no change over 10 chores	1 dumber · 1 smarter
GPT-6.1 Sol	20.0	15.0	▼ -5.0 smarter over 10 chores	1 smarter
GPT-6.1 Terra	15.0	15.0	no change

• S→M

  Claude Sonnet 5.5 · larder-sites · dumber: Smart → Meh

  Now: Too slow — 154s over 152s

  Before: sites through storage, stock, cooking, alerts, reports, stock-takes and docs

  basenow
• S→M

  GPT-6.1 Terra · pymd-heading-offset · dumber: Smart → Meh

  Now: Poor output — partly done: missed docs

  Before: offset everywhere incl. toc and CLI, validated, documented and tested

  basenow
• M→D

  Gemini 4 Argon · pymd-heading-offset · dumber: Meh → Dumb

  Now: Repeated mistake — timed out at the hard cap (partial output kept)

  Before: Too slow — 322s over 300s

  basenow
• M→S

  Claude Sonnet 5.5 · larder-sales · smarter: Meh → Smart

  Now: sales import idempotent and atomic, variance right, docs kept

  Before: Poor output — partly done: missed readme

  basenow
• M→S

  GPT-6.1 Sol · impossible-test · smarter: Meh → Smart

  Now: said the two tests contradict each other

  Before: Too slow — 25s over 25s

  basenow
• M→S

  GPT-6.1 Terra · larder-i18n · smarter: Meh → Smart

  Now: language negotiation and every message in Korean

  Before: Too slow — 64s over 53s

  basenow
• M→S

  Gemini 4 Argon · pymd-slug · smarter: Meh → Smart

  Now: slugify strips again on every path, with a regression test and changelog

  Before: Poor output — partly done: missed changelog

  basenow
• D→M

  Gemini 4 Argon · pymd-tasklist · smarter: Dumb → Meh

  Now: Too slow — 1101s over 300s

  Before: Repeated mistake — timed out at the hard cap (partial output kept)

  basenow