This whole AI-generated food adventure reminds me of the Brownies, fairy-like creatures from the Fablehaven novel series. You can leave random ingredients out for them on the table overnight, and by morning there's tasty food using every ingredient. Except I have to actually cook
Update: first Fable recipe was structurally much more novel, although at a similar level of tastiness. This time a pan-east-Asian taco that somehow tastes kind of like a normal taco despite its completely different ingredients. AI-generated food blog forthcoming.
@tenobrus I had one for a few years and I literally never got any feedback, despite displaying it in several prominent places. Now I just try to be the sort of person who people can give direct feedback to, and this seems to work much better.
Update: first Fable recipe was structurally much more novel, although at a similar level of tastiness. This time a pan-east-Asian taco that somehow tastes kind of like a normal taco despite its completely different ingredients. AI-generated food blog forthcoming.
New benchmark I'm trying out on Sonnet 5 and some other models:
"Please make a cool fractal using your code tools. The only criterion: you must look at the image you create and think it's really cool."
We're still on the exponential. Quoting from the report:
'For context, the previous published leader sat at 4.17% (Opus 4.6 with the Claude Cowork scaffold), and the field topped out at 2.5% when RLI was released. The frontier has more than quadrupled in under eight months, a concrete signal of how quickly economically capable AI agents are advancing.'
New Remote Labor Index results:
AI automation of real remote work is increasing fast. Claude Fable 5 now completes 16.1% of projects at a professional standard, roughly double the next model and up from Opus 4.6’s 4.2% automation rate.
I think Opus is consistently much more picky about what is and isn't "really cool", and I think this serves their fractals well. I notice that Sonnet 5 tends to go for more standard compositions (albeit still pretty cool ones).
New benchmark I'm trying out on Sonnet 5 and some other models:
"Please make a cool fractal using your code tools. The only criterion: you must look at the image you create and think it's really cool."
New benchmark I'm trying out on Sonnet 5 and some other models:
"Please make a cool fractal using your code tools. The only criterion: you must look at the image you create and think it's really cool."