AI is starting to run engineering simulation. Someone independent has to check it.

We are building the independent check for AI-run engineering.

0 of 23
wrong results accepted
BENCHMARKS §1 · verdicts.json
20 of 20
wrong-problem runs caught against the task
BENCHMARKS §3b · rm18.json
0 of 28
copied answers accepted
BENCHMARKS §3c · silentbench.json

Team.

Updates.

  1. Released: courant 0.5.0, open source under Apache-2.0.The release
  2. The premise proved: of the wrong-problem runs, 20 of 20 caught against the task and 8 of 20 without it.The evidence

What we say, and how.