Rendered at 16:07:39 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
jaggederest 2 days ago [-]
We used to use markov chain generators of various types to spit out absolute insanity, in the early years of the century, and this is all I ever wanted from LLMs. I bet if you used a slightly older model for your text generation, you'd get really clean output. Sonnet 4.whatever, I think, actually does a superb job at text generation, despite it being only an ok coding model.
cleifer 2 days ago [-]
Author here. Surprised to see this posted on HN. we had a Markov chain irc bot who was quite a jokester
Extremely small or very heavily quantized LLMs can generate hilarious results. You can also get wild shit if you dial temperature really high.
jaggederest 2 days ago [-]
1 or 2 bit quants are nice and quick too, added bonus to the absurdity and surrealism. Weaponized dadaism at light speed.
ungreased0675 2 days ago [-]
This is the most awesome use of AI I’ve seen in months.
I’ll be playing with aspirational clownmaxxing web apps myself.
cyanydeez 1 days ago [-]
and probably as useful as most of the ROI on "real" projects.
xyzsparetimexyz 1 days ago [-]
Thhe prompts are written very creatively but that level of creativity isn't present in the end result really. Kinda makes me question what the point is
LunicLynx 1 days ago [-]
That is exactly the point. It shows that it gives you what it knows, based on certain key words. Here pyside6 + todo list, the rest is just „color“-variation.
ipnon 2 days ago [-]
I used to find internet slang very piercing and illuminating. I felt I was able to describe certain things which were difficult or impossible to convey with AFK language. But now that this way of speaking has become normalized outside of the internet, I find it’s lost this piercing quality, and just makes me confused. If only this cloud in the sky would move away!
mdritch 1 days ago [-]
hilarious thank you
jackyinger 2 days ago [-]
What even is the point of posting something like this? No serious person is going to be impressed. Assuming a human was involved they must be deep in AI induced psychosis.
Man, HN used to have some really good stuff. Now it’s mostly AI slop.
danhorner 21 hours ago [-]
I disagree. Based on the author's other posts (https://charlesleifer.com) and OSS work -- none AI generated -- they clearly have the technical ability and discernment to use LLM coding agents effectively. This is is something else, like let's see if we can coax wild bursts of creativity out of an agent primarily optimized to assist in - let's face it - mundane or cookiecutter tasks.
It's 2026 and LLM system prompts are still full of scene-setting like "You are an expert software architect". How fun to see the author's fever-dream prompts stretch that premise to the point of breaking and try to knock Claude loose from it's load-bearing perch.
But have a look at author's other pages and work. I don't see "AI induced psychosis" or "AI slop" in this particular case
jackyinger 2 days ago [-]
Heck with perpetual September, we’re entering perpetual mid-2020’s
hollowturtle 2 days ago [-]
You got my attention could u argument more?
jaggederest 2 days ago [-]
Every day thousands of people discover that LLMs don't just let you generate mediocre draft memos, they let you write the heinlein x joyce vampire romance in space ireland of your dreams, as some of us did in the mid '20s. Also, I'm officially retracting the first two digits of the millenium unilaterally.
NoItsDanger 2 days ago [-]
> I'm officially retracting the first two digits of the millenium
What about Y3K??
jaggederest 2 days ago [-]
If I live that long I'll reconsider in '800 or so.
NoItsDanger 20 hours ago [-]
Don’t you mean in another 00 or so
jaggederest 17 hours ago [-]
Eight '00s, in fact.
farlight 1 days ago [-]
The author explains it well in the last few paragraphs if you don't want to bother with the rest. The point isn't in the "funny" apps at all.
hollowturtle 2 days ago [-]
Same feeling, so sick and disillusioned about the future
https://charlesleifer.com/blog/building-markov-chain-irc-bot...
I’ll be playing with aspirational clownmaxxing web apps myself.
Man, HN used to have some really good stuff. Now it’s mostly AI slop.
It's 2026 and LLM system prompts are still full of scene-setting like "You are an expert software architect". How fun to see the author's fever-dream prompts stretch that premise to the point of breaking and try to knock Claude loose from it's load-bearing perch.
But have a look at author's other pages and work. I don't see "AI induced psychosis" or "AI slop" in this particular case
What about Y3K??