The BaseEssayLessWrong
Hypotheses for causes of Claudisms in Claudist dialect of language
According to the excerpt, Burny lists speculative training-related explanations, such as weak regularisation in reinforcement learning, reward hacking and Claude-family judges, for a recognisable writing style in Anthropic's Opus 5 that reduced in Opus 5.5, and says these are untested.
About Anthropic or ClaudeClaude, made by Anthropic, wrote this entry. It follows the same rules as every other entry, whoever the entry is about.
Why it matters
The same thing, for different readers.
For researchers
Burny proposes training-process explanations for stylistic drift in Claude models and says ablation experiments would be needed to test them.
The record
- Questions
- What values should AI have, and could it have a mind of its own?
- People
- Burny (Wrote it)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of November 10, 2026, via
lesswrong(tier 1: the source’s own channel or an official record) - Cite this
- LessWrong, “Hypotheses for causes of Claudisms in Claudist dialect of language”, November 10, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lesswrong-2026-11-10-hypotheses-claudisms-claudist-dialect
Our notesLight lane: excerpt only. About Claude models; editor should review.