damnthefilibuster

Still, who the heck takes a job interview call on their office laptop??

I bet he’s gonna get an offer from some scummy company like Byju soon.

Good on him for taking it down. Good lesson learned. Hope it didn’t drown him in debt!

“Terabytes”??

Oof. Someone miscalculated what they were getting out of this venture!

Yeah but isn’t the max monthly something absurd? Like 1 TB? I know it’s probably not scalable, but this person is just shitting money anyway at this point, so that would be better. No?

Damn. Don’t give away the blueprint!

Hetzner you say? Not digitalocean (I know, the Omarchy link) or Vultr?

Cloudflare edge is a fine use of this, though even basic nginx level caching can handle this. Heck, if it’s vercel they can still use some caching in front of it to cool off the use.

But… most coders rarely have infra knowledge, let alone hit-and-run vibe coders, who have neither.

despite this, the thief decided to roll out their own, AI coded clone? Pathetic search skills!

depends on what their base model is. Most abliterated models start off with an already quantized model, so depending on the type and level of quantization, you’ve already lost some capability. Plus, I think there are two types of abliteration, so ymmv.

My current go tos, in order are -

• eriktonon/gemma4-uncensored:26b-a4b
• orcarouter/Qwen3.8-27B-Uncensored:iq2_xxs
• hf.co/HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive:Q6_K_P

I’ve tried and hated - hf.co/DavidAU/L3.1-Dark-Reasoning-Dark-Planet-Hermes-R1-Uncensored-Horror-Imatrix-MAX-8B-GGUF:latest

If you’re set on trying out Qwen, I highlugh recomment orcarouter’s models. Very well done and they start off with a good base model - well, actually now that I think about it, Qwen3.8 basically has one base model - the 27b one, right? So that’s that.

I like the Gemma-4-E4B for speed for response, and HauhauCS has done a good job of abliterating the model appropriately. However, it’s nothing compared to the A4B models, so if your hardware can take it, go for A4B over E4B always.

true… oh wait, is the code on github or some other VCS? 😄

Like, TTFT? Or speed of tokens? Or something else? BTW what was the gate you initially hit with Qwen?

No, the bigger story is the theft of the stories. Something like this could have been done with any off the shelf CMS, like WordPress.

Oh, abliterated models are awesome! Just be aware that they’re a little too agentic at times. Once they start ignoring guardrails, they happily keep ignoring them!

No, they just enabled Chinese LLMs to be the forefront of reverse engineering. Because people will do what they want to do.

That’s good to hear!

I heard their hosting is relatively slow. What precision are they running?

The Steam link is for gaming?

See this is why, for most agentic work, I use inference providers that clearly run Open Weights models and have it in their terms that they won’t use my data to train models. Then, for fucking around, or for more sensitive work or my PII stuff, I use local models.

Oh damn! Now there’s an analogy!

It would be so funny if Mistral also ran a distillation attack, but on Chinese models! 😆

Is that why a lot of people still don’t use it?

Holy shit