HN user
It's been known for several years that LLM activations encode future tokens ahead of time (e.g. https://arxiv.org/abs/2404.00859).
But this has only been shown on simple tasks, so I think this paper is still quite neat. The interesting thing is that they show "future horizon length" varies across models.
Considering the job description is unrelated to the company and the title, this is a scam post, not quite a "ghost job".
They're probably rapidly opening + closing new jobs to increase visibility, as matching models on job boards tend to prioritize new posts.
Given their pricing, I'd guess their models are just way bigger in parameter count. They've always underperformed in cost-per-performance.
They also target a cost-insensitive market (corporate/coding users) compared to Google/OpenAI which support massive amounts of free users.
Love your work on this, thanks for bringing the ggplot syntax to Python!
Yeah, the footnote/sidenote on the paper (the one labeled #2) mentions this as well so you can't type that directly
I believe they are trained for security now, but you're not wrong in that it's kind of stapled on top
A lot of open weight models don't understand intent well, they'll overfixate on a word in the prompt or just go off the rails trying to do much work.
GLM-5.2 actually has really good intent understanding though, on par with GPT-5.5 and Opus from my experience.
This works because Nex itself is a finetune of Qwen3.5 (https://huggingface.co/nex-agi/Nex-N2-Pro). It's merging Qwen3.5 with a Qwen3.5 finetune.
I don't believe this would work on two LLMs that have different pretraining. Even if it did you would need two LLMs that have exact same internal activation shapes, dimensions, expert counts, token vocabulary, realistically it would never happen outside of finetunes or academic experiments.
True, but coding agents have only been practically useful since early 2025. Jobs are flat or a bit up since then.
This paper has an major issue that they are not surfacing, these activations can just be correlated on a common latent. For example, both the original activation and the explanation could share a broad latent like "this is an adversarial scenario". That could make reconstruction loss look good without showing that the actual explanation was the correct cause for the LLM's response.
I find this rather disturbing. Anthropic has quite a habit of overclaiming on questionable research results when they definitely know better. For example, their linked circuits blogpost ("The Biology of LLMs") was released after these methods were known to have major credibility issues in the field (e.g., see this from Deepmind - https://www.lesswrong.com/posts/4uXCAJNuPKtKBsi28/negative-r...). Similarly this new blog is heavily based on another academic paper (LatentQA) and the correlation/causation issue is already known.
Shoddy methodology is whatever, but it feels like this is always been done intentionally with the goal of trying to humanize LLMs or overhype their similarities to biological entities. What is the agenda here?