I also think he says this because nothing he has ever shipped has had any real substance.
It was slop before AI slop existed.
Which, to each their own, he was very successful doing so and found an avenue that worked during those times. Its just it doesn't necessarily reflect onto everything else.
It's wild how polarized opinions are on the exact same AI models right now.
One person says Fable feels stable and productive, the next calls it awful and painfully token heavy.
Same thing happening with GPT models.
Anyone have opinions on why this is happening?
Are we all deep in AI psychosis? π΅βπ«
@HealthRanger I would be petrified if I was an Anthropic investor...
The only thing they are hanging on by is a good model... That they also can't even handle compute for.
OpenAI, SpaceX, Google, even Zuck are positioned much better than they are.
The new "Plan" mode changes feel like many steps backward.
Timing out the questions instead of waiting for my response is not useful. (I'll send off a large prompt and then begin doing something else, and when I come back it's not waiting for my input. It either kills the run or just skips it.)
Not even bothering to ask questions on some plans as a whole.
The whole point of "Plan" is trying to narrow the scope of what I want for the output.
I know the goal is to move towards more hands-off, but that really defeats the purpose of specifically the Plan mode itself.
It's a hard pill to swallow for those of us who have spent years learning and enjoying the art.
But I feel like if I don't willingly let my coding muscle atrophy, it will ironically just cause me to fall behind.
It's very hard to deny that coding will be mostly hands-off in a matter of time (and many would argue that time is now anyway).
@StringHemu The only reason would be Fable. The new 5.6 models for me are just not doing it in terms of quality at all. Codex has been great especially in the Opus vs 5.5 days, but Fable is a real machine. My big tasks are all going to Claude currently.