czeizel

the ability to let fable work independently, run experiments and only check in occasionally is probably claude's strongest argument here. i'm also surprised by how codex-heavy the replies have been, although the pricing and usage limits seem to explain part of it.

what kind of tasks are you comfortable leaving fable alone with, and how often does it take the project in a direction you need to undo? also, gpt translating the pro-claude comment is a nice ending :)

interesting, that probably shows how subjective "humanlike" is here! thx for your reply!

what makes it feel more humanlike to you - the way it reasons through problems, the tone of the responses, or is it simply more comfortable to work with day to day?