C.W.K.
Stream
Lesson 10 of 10 · published

Debugging Bad Reasoning — Read the Trace

~14 min · reasoning, debugging

Level 0Apprentice
0 XP0/100 lessons0/14 achievements
0/120 XP to next level120 XP to go0% complete

The trace tells you where it went wrong

When a reasoning model produces a wrong answer, the trace is the evidence. Most teams never read the trace — they tweak the prompt by feel, ship a fix that may or may not work, and move on. Reading the trace is slower up front and faster across a quarter.

What to look for

  • Wrong premise — model anchored on a phrase from the input that wasn't load-bearing.
  • Skipped step — model jumped to a conclusion without checking a constraint.
  • Repeated step — model re-stated the same sub-result without progress.
  • Verifier rejection — model proposed a plan that the verifier killed; the next branch wasn't materially different.
  • Token-budget exhaustion — trace ran out of room before the answer.

Trace as eval input

Once you have a few annotated bad traces, the failure modes become test cases. Add them to your eval set. The next prompt change that fixes them counts as progress; one that doesn't, doesn't.

Code

Trace logging shape·json
{
  "request_id": "req_abc123",
  "model": "claude-opus-4-7",
  "prompt_version": "support.draft@7",
  "thinking_tokens": 4823,
  "output_tokens": 287,
  "thinking": "...",
  "answer": "...",
  "verifier_pass": false,
  "verifier_error": "missing required field 'sla_tier'"
}

External links

Exercise

Pull 10 failed-verifier responses from the last week. Read each trace. Classify the failure mode. Identify whether the prompt or the verifier is the right place to fix it.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.