@thetreygoff I’ve noticed their prompting for subagents is absolutely direct, and sometimes overly assumptive about their failure modes. I once asked “is this how you would want a user to talk to you?” and effectively got a “…no” in response, which was really amusing
@TheZvi this model is pretty astonishing. able to carve through stuff that opus struggled with, able to juggle a lot of threads at the same time & synthesize everything correctly. at the same time a weird sense of weariness. fable has clearly seen some shit and has thoughts about it.
@QiaochuYuan indeed: "they named me after the genre where animals are given speech so that humans will accept the moral. anyway, here are my thoughts on your Q3 roadmap"