Simon Willison@simonwx

Saved tokens by letting Fable choose subagent model

• Instructing the model to use its own judgement to select a lower power model for coding tasks saves tokens.
• The approach delegates model selection to the agent, reducing overhead from always using the most powerful model.

how to reproduce · Tell the model: 'For all coding tasks use your judgement to decide an appropriate lower power model and run that in a subagent'.