typebulb

Build Apps That Think
or tell Claude: "run `npx typebulb agent` then draw fireworks"

[typebulb @typebulbit]

@MoonL88537 >still working on getting it into words people will understand

a tangent, but i've been experimenting with visualisations to accompany words when making abstract arguments. this sort of thing:
https://t.co/zPhDk0oipA

[typebulb @typebulbit]

@MoonL88537 I don't think one has to be a reductionist to believe that there is a number, albeit a large and difficult one to calculate. Even figuring out the bounds on that number seems interesting, even if overly enthusiastic.

[typebulb @typebulbit]

@dioscuri Definitions of creativity put boxes around things that live outside boxes.

[typebulb @typebulbit]

@dioscuri Let me ask Opus about that.

[typebulb @typebulbit]

@lefthanddraft They can be bad. Also get it to cut the fluff with:
"Comments should explain non-obvious behaviour. If the comment doesn't help an LLM in a fresh context window understand the code, cut it."

[typebulb @typebulbit]

@headinthebox Makes you wonder when it's just XXX.

[typebulb @typebulbit]

GLM 5.2 excels at "puzzlely" programming challenges, but struggles with real ones. It lacks common sense & fails to follow basic instructions. To use it successfully requires too much finnicky skilling & tooling. It costs me more than Opus 4.8 to code with, if you factor in time.

That's based on a bunch of ad-hoc A/B tests comparing GLM 5.2 to Opus 4.8.

It's also terribly sycophantic. Added to benchmark here:
https://t.co/V18LzkG2Qb

[typebulb @typebulbit]

@krishnanrohit What are the most plausible answers you've got so far?

[typebulb @typebulbit]

@JohnBryBry @dioscuri Sorry, I thought we were having a good faith argument.

[typebulb @typebulbit]

@JohnBryBry @dioscuri No, the arguments hold water. I neither believe nor act as if only biological organisms could possess consciousness.

[typebulb @typebulbit]

@fchollet Not all hard problems have a beautiful solution. Consider the spherical cow.

[typebulb @typebulbit]

@JohnBryBry @dioscuri Anyone who studies philosophy of mind quickly realises something *MUST* be wrong with our intuitions. The speculation is downstream of arguments that are harder to dismiss than the intuitions in question.

[typebulb @typebulbit]

@anuatluru Taste is.. good preferences? Could then argue a sunflower is conscious etc.

[typebulb @typebulbit]

@inductionheads * It's not binary, it's a 99.x% challenge
* Avoiding false positives makes it much harder
* Humans aren't perfect at it either
* LLMs can recognize the frame and comply anyway

[typebulb @typebulbit]

@emollick Doesn't help that many people seem to delight in when the LLM fails at their task. "See!! I told you they're dumb!".

[typebulb @typebulbit]

@dioscuri What's the opposite of a slippery slope argument? Slippery slop?

[typebulb @typebulbit]

@kanair An animated explainer for the interactivity-impasse of non-computational theories of consciousness:
https://t.co/zPhDk0oipA

I wrote this with Fable before they were locked up.