New pattern this week: Fable at full force, maxed out, orchestrating ChatGPT 5.5.
The logic is tokens. The ChatGPT subscription offers vast quantities of them. Fable’s are scarcer and dearer, so the sensible split is to spend Fable on judgment and ChatGPT on volume. A wrong plan executed at speed is the most expensive thing a cheap model can produce. Making sure that each ChatGPT builder does the right job, the right way, can be Fable’s job.
Some context on the two halves. GPT-5.5 arrived on April 23 for paid ChatGPT plans. Back in April it ran for 13.4 hours flat for me, invoking Claude and Gemini when I’d asked to review Codex’s own work, without stopping. It has the stamina. What a builder with that much stamina needs is a foreman.
Fable is two days old. Anthropic’s pitch is that it “can work autonomously for longer than any previous Claude models”, with its lead growing as tasks get longer and more complex, and through June 22 it’s included on paid Claude plans at no extra cost. On the API it runs $10 per million input tokens and $50 per million output. ChatGPT Pro, meanwhile, has been $200 a month since OpenAI introduced it in December 2024. Those two prices, side by side, are the whole argument.
It should be reading the plan, checking the work and sending it back when it’s wrong. That’s an orchestrator’s job, and this week it’s Fable’s.
This is lending me far more bang for my buck right now.
