[Model behavior] Long answers after repeated explicit brevity instructions; the user had to interrupt mid-answer several times (recurrence of #99079)
Status Open
Maintainer reply None cached
Activity 0 comments · opened Oct 14, 2026
What happened
The loaded instructions require the verdict first and one to three lines when nothing is wrong. The model repeatedly wrote multi-paragraph answers, including for simple yes/no questions and clean verdicts. The user had to interrupt several answers part-way and ask for the simple answer.
Expected
Verdict or answer first, short, with detail only for blockers.
Impact
The user spends time reading and interrupting instead of deciding.