hidden context costs

1 articles · 4 co-occurring · 0 contradictions · 0 briefs

Thinking tokens represent a form of hidden context cost: they're consumed within the context window but don't produce useful output, only internal deliberation loops. This is a context efficiency fail

Related concepts

token efficiency optimization 1 prompt engineering as band aid 1 model selection strategy 1 inference optimization 1

Evidence chain (1 articles, showing 1)

@IntuitMachine: GLM 5.2 trades off smaller LLM size for a lot more thinking tokens. It's a g... example_of

query this concept

$ db.articles("hidden-context-costs")

$ db.cooccurrence("hidden-context-costs")

$ db.contradictions("hidden-context-costs")