All workflows

Workflow runs

• Fall back when the pinned model is retired, and say so

  completed successfully:
  ci#10:Commitcc0066dpushedbypcbeingused333
  main

  Sep 1, 20261m 37s
  More actions for workflow run 10

• Publish the first clean answers run, and test the judge that produced it

  completed successfully:
  ci#9:Commit1a656b2pushedbypcbeingused333
  main

  Sep 1, 20261m 29s
  More actions for workflow run 9

• fix(evals): read a judge that inlines its reasoning

  completed successfully:
  ci#8:Commit1ccd89cpushedbypcbeingused333
  main

  Aug 26, 20261m 27s
  More actions for workflow run 8

• Stop the judge grading its own homework

  completed successfully:
  ci#7:Commit9c719d2pushedbypcbeingused333
  main

  Aug 25, 20261m 37s
  More actions for workflow run 7

• Point the model override comment at the measurement

  completed successfully:
  ci#6:Commit5dd38fapushedbypcbeingused333
  main

  Aug 25, 20261m 43s
  More actions for workflow run 6

• Rerun the pending evals, and fix what the runs broke on

  cancelled:
  ci#5:Commit1a547e6pushedbypcbeingused333
  main

  Aug 25, 20261m 29s
  More actions for workflow run 5

• Cut the cold-start memory peak that was risking the public demo

  completed successfully:
  ci#4:Commit8452584pushedbypcbeingused333
  main

  Aug 13, 20261m 18s
  More actions for workflow run 4

• Rewrite the README around what the new numbers say

  completed successfully:
  ci#3:Commitcf9667apushedbypcbeingused333
  main

  Aug 13, 20261m 30s
  More actions for workflow run 3

• Make the suite pass under bare pytest, not just python -m pytest

  completed successfully:
  ci#2:Commit45388b3pushedbypcbeingused333
  main

  Aug 12, 20261m 43s
  More actions for workflow run 2

• Run the tests on every push, and test the app itself

  failed:
  ci#1:Commitaad216fpushedbypcbeingused333
  main

  Aug 12, 20261m 29s
  More actions for workflow run 1