[Bug] Consistently, confusing new phrases and fillers are being unavoidably inserted

Status Open
Reported on v2.1.247
Maintainer reply None cached
Activity 0 comments · opened Aug 27, 2026

Bug Description
Has regressed into making up words that cause confusion and slow me down immensely. Claude, please check if your AI might have fallen into doing this because of some inbuilt attempt to cut down how often users can ask it questions. It might have decided that confusing us is a good way to reduce compute power so that we slow down. Prompts instructing to be concise, to just state material or not to interpret etc are not helping. See example below:

CC: One thing worth deciding now rather than later: the shell landscape's P700 offset is −149.2, which already lowers the special pair. S_RC then sets how much extra stabilisation the RC exciton gets from the bath on top of that. At S = 2.6 those two act together, which is part of why the shell runs came out fast. The S = 1 arm will show how much of the 42.4 ps depends on the Huang–Rhys factor rather than on the landscape — which is exactly the anchor you wanted, just with the labels the other way round."

Me : What "deciding"? whats the decision i have to make there?

CC: None — that was filler. The primary/anchor labelling was already settled in your instruction (S = 1 primary, 2.6 as the anchor for whenever someone measures the P700 Huang–Rhys factor), and both arms are running. There's nothing for you to decide.

Environment Info

  • Platform: win32
  • Terminal: windows-terminal
  • Version: 2.1.247
  • Feedback ID: 5408f78e-0298-4bc1-96d7-feca683c4bab

Errors

[]

View original on GitHub ↗