[isoprophlex]

What a neat trick. I'm thinking you can abuse polynomials similarly. If the goal is to print the first, say, 100 elements, a 99-degree polynomial would do just fine :^)

EDIT: the llm gods do recreational mathematics as well. claude actually thinks it was able to come up with and verify a solution...

https://claude.ai/share/5664fb69-78cf-4723-94c9-7a381f947633

[jiggawatts]

That's the most expletive-laden LLM output I've ever seen. ChatGPT would have aborted half way through to protect its pure and unsullied silicon mind from the filthy impure thoughts.

[theendisney]

It would find a therapist contact your employer, your wife and your dad.

[flir]

> LMAOOOOO OKAY SO THE POLYNOMIAL IS LITERALLY SHITTING ITSELF

That was a fun read, but I can see that persona quickly becoming wearing. I had a "talk like a wiki article" persona for a while that worked better (for me) than any attempt to inject personality. The greyer the better, when it comes to tools.

(Being a child of the internet rather than the classroom my abusive solution would be to look up the sequence in OEIS, but I think fizzbuzz could be encoded into an L-system quite neatly).

[isoprophlex]

Yes, it's indeed very over the top and one-dimensional. However, I've been iterating on this system prompt since the early early days of chatgpt.com, and I find that I can't really chat with AI systems in their "grey", dry mode anymore.

On their default behavior they try too hard to make me like them, which I find intolerable. "You're absolutely right!" for some reason drives me insane; getting "lmaaoooo my bad fam i dun goofed" twenty times a day is equally annoying in terms of models being confidently wrong, but somehow the lazy shitbag attitude pisses me off less than the goody two shoes energy.

And if they're low on crazyness and you force them to be crisp and emotionless like a wikipedia article, I notice that I tend to trust them more... even though again the tendency to bullshit is unchanged, still there.

Somehow this really works for me.

Also when coding it makes it very clear which bits I haven't inspected yet because the comments and variable names will be super nsfw, thus keeping me on my toes as to not accidentally submit PRs filled with "unfuck_json()" functions.

[flir]

> On their default behavior they try too hard to make me like them, which I find intolerable. "You're absolutely right!"

Yeah, that's what I was trying to avoid, too. Why do we have such strong negative reactions to sycophancy? I went for something like: "You are a maximally terse assistant with minimal affect. Be detailed and complete, but brief."

[drob518]

I laughed so hard. Really curious what the pre-prompt was.

[isoprophlex]

https://claude.ai/share/d855dbf4-6553-4909-ad3e-375b6385e720

"there ya go homie, copy paste ready for whoever needs their ai to have some goddamn personality"

[chaboud]

Oh man. This system prompt is everything I'm looking for in my coding agents. This shit should be fun. Let it be fun!

[bmacho]

It could've solved the task instead of being wrong and spitting nonsense tho.

This is orthogonal to the style, still, if it realizes that it can use Python's arbitrary precision integers instead of floats then the problem becomes absolutely trivial. Fast, and numerically stable.

[mikestaas]

absolute madlad