Fable 5 has been relaunched with much fanfare and unanimous praise, and for good reasons. My experience has been that, though very expensive, it is substantially smarter than all existing models. Discussing with Fable 5 felt like discussing with an extremely intelligent and meticulous person, who would agree with me if what I said was correct. The engaging conversations felt like real collaborations, where collaborators pool their intelligence to get the work done in the right way.
To do anything with Fable 5, I would first ask it to devise a comprehensive plan (which it sometimes can take up to half an hour to do, meticulously and without issues), and then use the following prompt to implement the plan:
Proceed to implement all the tasks in this plan in a new branch, and respect dependency requirements. Spawn subagents (in parallel, if possible) to implement each task. For all subagents that are writing related, including writing parts of a paper or a plan, always use Fable 5 at high effort. For subagents that are coding or review related, always choose Opus 4.8 at xhigh effort or GPT 5.5 xhigh using the codex plugin. For each task, spawn its own fresh subagent to implement it with well-defined acceptance criteria, and then spawn another subagent to review the implementation and whether the criteria is satisfied. When a review subagent has findings that are of critical, high, or medium priority that you believe should be fixed, reuse the existing implementation subagent to fix them when they’re still available (instead of spawning fresh ones). Also, reuse the existing review subagent to review again. Iterate through the fix-review cycle until the review subagent has no more findings of critical, high, or medium priority and that do not add more features or over-engineer. If you happen to divide the tasks into phases, then add a review gate, also as a subagent, to review each phase before committing. After completion, each phase or task should be committed with a detailed commit message, and pushed to the remote origin. When all the phases or tasks have been completed, PR the entire revision with a detailed description.
I also have the Codex Claude plugin installed; but sometimes a Codex agent launched by this plugin may get stuck doing nothing. I am not sure why.