Skip to content

AI Anti-pattern

The Correction Spiral

Burning turns on corrections instead of resetting context.

Why it matters

Every correction turn consumes context window and rarely yields the fix. After two failed attempts, the model's state is the problem, not the wording. A new conversation costs nothing and almost always resolves it faster.

  • Anthropic Claude Code guidance: reset after two failed correction attempts
  • Correction spirals are a documented failure mode in extended-context reasoning tasks

What this looks like

Three or more consecutive correction turns on the same point in a single conversation, or multiple threads that end on an unresolved blocker rather than a clean pivot.

This is a generic example. Run your own analysis to see instances from your actual AI conversations.

See this in your own conversations

Paste a conversation with Claude, ChatGPT, or Gemini to find out if you show signs of The Correction Spiral.

Run your own analysis →