What's this about?
A budgeting technique. In `/usage` you see two bars: your weekly total and your Fable-specific usage. Fable is smarter and more expensive, and its allowance is only 50% of the total, so two failure modes appear. Fable at 70% while total is at 40% means your intelligence budget runs out and you spend the rest of the week on weaker models. Total at 70% while Fable is at 20% means cheap models burned the budget and the strongest model's window went to waste. The fix is not a mechanical half-and-half but keeping the two bars descending in roughly the proportion that suits you: main session on Fable for judgement, architecture, task decomposition and review; execution, retrieval, file edits and test runs dispatched to Opus subagents. `/clear` on task switches, `/compact` mid-task, and check the bars a few times a day.
Why is this a real problem rather than fussiness?
Because the two meters are independent and neither is the binding one by default. Most people track only the total, discover Fable is exhausted on Wednesday, and spend the rest of the week at reduced capability without having decided to. Making it visible turns an accident into a choice.
Why does this matter to you?
- It converts the week's quota complaints into something actionable rather than something to be annoyed about.
- The proportional rule is better advice than any fixed ratio, because usage shapes differ.
- Pairs with `/doctor` and the 1M-context setting: three independent levers on the same budget.