I had GLM-5.3 in OpenCode make this while I drank Guiness 3 through 8. Then I posted it to HN before bed, thinking a couple other people might get a kick out of it.
~250k requests and a brief stint at #1 later, and I'm amazed that my 2 RTX-4060ti's haven't collapsed under the load. The model is an abliterated version of Qwen3.6 35B running on my PC. (qwen3.6-35b-a3b-uncensored-hauhaucs-aggressive)
So anyway, it's just a fun satirical sort of protest directed at the absurdity of it all. No attempts to monetize have been made or will be made. I'm already at 76% of my daily quota for the cloudflare workers, so it won't be around much longer. My budget for this project was the $10 I spent on the domain.
As a side note, that particular model is extremely abliterated, although it's almost certainly wrong, it'll attempt to give you an entire setup plan for a meth lab if you ask it. Or something similarly naughty as a test.
This sort of absurdist satire with the right system prompt is a near perfect use case for it since it'll happily cooperate with just about anything. And yet it's big enough to be able to write coherently and provide a reasonable answer.
I think it scores like 0 out of 460 on a test of known "bad" prompts that most LLM will refuse.
34679 · · focus · HN ↗
~250k requests and a brief stint at #1 later, and I'm amazed that my 2 RTX-4060ti's haven't collapsed under the load. The model is an abliterated version of Qwen3.6 35B running on my PC. (qwen3.6-35b-a3b-uncensored-hauhaucs-aggressive)
So anyway, it's just a fun satirical sort of protest directed at the absurdity of it all. No attempts to monetize have been made or will be made. I'm already at 76% of my daily quota for the cloudflare workers, so it won't be around much longer. My budget for this project was the $10 I spent on the domain.
walrus01 · · focus · HN ↗
This sort of absurdist satire with the right system prompt is a near perfect use case for it since it'll happily cooperate with just about anything. And yet it's big enough to be able to write coherently and provide a reasonable answer.
I think it scores like 0 out of 460 on a test of known "bad" prompts that most LLM will refuse.