Wasn’t all that obvious to anyone who read the original post when they got to the section that explicitly explained what the author of this post is now summarizing?

Sure if you only read the title and brief at the top you’d walk away without the full picture though.

Yeah. The article is not misleading. The title is, but you know, headlines are all lies that's why you read the article.

The title is: "LLMs can natively read & write neural network weights"

If you read comments about the paper it's pretty obvious people are misled. People think that the method read in Qwen weights directly to improve it. That's not what it did --- it authored a program to analyse and fix the weights.

What's confusing is they do have experiments that read raw weights, and the abstract is misleadingly written to make it sound like that's how the Qwen experiment works.

I wouldn't say misled, I would just say they made up their mind about what it means regardless of what the article said. Maybe the title is misleading, but the original article is pretty fair.

> who read the original post

I think there's your answer. A hell of a lot of people will often read the title without reading the contents of the article. I admit I'm often guilty of that

At this moment on the front page (early morning in the US), this post is #1 and the other one is #2, so it's working like a warning about the other one being misleading before clicking on it.

[delayed]

Wow that was fast

The original post explains what they did and why better than this post. This post taught me nothing that wasn’t in the original.

They make it sound like they improve Qwen by reading raw weights. They don't. In fact the experiments that use raw weights have very little to do with the experiments that work on larger models.

The paper makes it ridiculously clear what they do and don’t do. And they make the reasons why clear too. There is absolutely nothing interesting in this new article.

I don't understand how anyone can read the abstract and think it's "ridiculously clear" that the "Just by looking at the raw weights," doesn't apply to the Qwen experiment.

It says this “All of this is possible without allowing Astra to write training loops. It uses its reasoning to figure out what each part of a neural network computes, or should compute, and changes the weights accordingly.”

This seems 100% accurate to me. What is a lie here?

Nobody said they outright “lied”, most misleading statements are “100% accurate” yet are designed to mislead the reader.

And when you see the reactions in the HN comment to the original article, it's pretty clear that most people have indeed been mislead.

They did, except they also did write an abstract and a title that will mislead anyone who isn't going to read the full paper in details.

It's fair to say that this is malpractice.

I just reread both and neither seem misleading to me. If anything the title is underwhelming.

> The “just by looking at the raw weights” setting (raw floats in context, no tools, no running the network) only appears in experiments 1, 2 and 4, on networks of 17K–172K weights.

The author is being equally disingenuous.

That piece by itself is insightful. Meaning, LLMs are capable of deciphering what a model can do by looking at the weights alone.

The fact that we can’t squeeze enough of them through the context is a technical limitation. However, this does not diminish the importance of this finding/discovery. This is an unexpected emergent behaviour.

Yeah, I can read the weights in a small neural network, and for a given input, given that I know the activation functions, I can figure out what it does with pen and paper and lots of spare time.

I doubt this is either unexpected or emergent behavior. Anthropic and OpenAI are explicitly racing to reach RSI and I would expect them both to have trained their models on this problem, in hopes that it would generalize.

If there's a linear relationship between the weights and the output, it's not so hard to read the weights and reason about what's happening. If you read the reasoning traces for how the models do those tasks, you'll see that they rely on those direct relationships.

The same could never work in a larger network. It's not about just capacity to feed in the weights, it's about the complexity of the relationship. There's no demonstration here that an LLM can do that, and it doesn't really make sense to want it to. It makes much more sense to author code to analyse and modify than it does to do it directly.

I dunno, I'm still really impressed by the "Sparks of Weight-Space Programming" idea. Perhaps its because I didn't know that LLMs could do understand or successfully modify weights at all previous to reading it. However, I think the examples provided in the original piece were eye-opening. I'm sure humans can build tooling to also modify weights, but the author here agrees that it's tough. The examples where Astra was modifying raw weights are still very impressive to me.

Here's the work that shows how rule-based logic can be compiled into transformer weights:

* Thinking Like Transformers” by Weiss, Goldberg & Yahav (ICML 2021).: https://arxiv.org/abs/2106.06981 * Tracr: Compiled Transformers as a Laboratory for Interpretability https://arxiv.org/abs/2301.05062 (code at https://github.com/deepmind/tracr)

So it's not really an "LLMs can modify weights" thing. Rather it's "LLMs can write programs that can write or modify weights". Indeed LLMs can write many programs, it's impressive.

I feel like a lot of this Astra editing Qwen stuff is regurgitation of existing relatively well known techniques, for example: https://arxiv.org/html/2406.11717v3 and conceptually similar to: https://arxiv.org/abs/2512.05648 (tho this is done in training iirc)

Yes, while reading the blog I immediately thought about https://arxiv.org/html/2406.11717v3 paper. Original blog does actually overstates it findings as nothing like that exists.

[delayed]

That’s what I’m excited for: the raft of posts over the next few days as the inspired set to work

Agreed. Not sure it leads to anything truly groundbreaking, but more than likely will lead to very fascinating implementations that lead to even more interesting results 6-8 months from now. Think openclaw -> how people utilize now versus when it was first announced

A tangent, but I’m now realizing that the “recursive self improvement” thing that these LLM companies are talking about is (probably?) not about unsupervised Claude Code instances improving the training process, but rather LLMs modifying their own weights through tooling directly. It would certainly open up a whole new process for model optimization.

I was just watching a video about this. They had this survey question for AI researchers asking when they think AI would take their job. The reason they asked this was that would be the point where AI starts recursive self improvement. If it’s able to replace the job of an AI researcher, then it could keep improving itself.

This is a tautology.

Instead of just asking the researchers when RSI would be achieved? Sneeky!

Recursive self improvement permits recursion at any level of abstraction as long as k > 1.0 (much like supercriticality in nuclear reactions, which iirc is where the earlier term for RSI, “intelligence explosion”, comes from - substitute intelligence in for neutrons and it is the same math with similar apocalyptic implications).

It's obvious that recursive self-improvement has been happening and accelerating quickly for some time.

AI coding assistants make it much easier to conduct experiments and improve AI models, which in turn improves the coding assistants etc.

That's "just" improvement, not recursive self improvement.

oh man someone should make this into a loop where the model can alter it weights after an inference. some kind of descent of the local gradient.

Isn’t that what the Google dreaming thing is?

But can we use the idea inside the training loop, as an optimizer instead of (or next to) e.g. Adam?

something i found 15 minutes ago is that the car network test is "very very easy", as in Opus 5.5 in Claude Code did not even need to see the weights of the network to solve it correctly. i can give the transcript of the session. i'm pretty sure Opus didn't cheat

According to this person, some problems are only "solved" by modern models: https://nitter.netbub.com/AlexKrentsel/status/21173766837246...

(opus <= 4.8 and gpt <= 5.5 couldn't match a PPO baseline network, and 5.6-sol and opus5 could barely match it. The 20x baseline improvement only comes from opus5.5 and sol > 6.1 / Astra)

Welcome to publications involving ML in 2026... Glad to see someone else write this though because it's important to start having these discussions. Too much overselling has been happening for too long. There is so much great about ML. IMO it hurts the field to continue overselling claims like this.

I think the reviews are flawed for most of journals. But I can't really blame them, there is huge influx of new papers comming everyday. Unfortunately reviewers are more biased towards groundbreaking stuff so it seems they can't make connections with existing and prior work.

They are also biased to see from what institution you come from..

I think it's interesting how even inside tech, so many people seem to view neural network weights as "black magic" utterly beyond human comprehension. I guess that contributes to the sense of mystery around LLMs...

(And also the sharp difference in levels of sophistication you see inside neural nets and outside of it: Pipelines that feed text to LLMs so it can detect highly abstract and nuanced semantic differences in the text's contents - but the pipeline still uses regexes for sentence splitting...)

A better idea is to have Astra modify it's own weights and let it try if it can RSI, but you gonna need to closely monitor for actual leap in capabilities so you can keep a check on alignment.

Someone has to already be doing this right?

Given the sizes of the models and the cost of the supercomputer necessary to run such an experiment, I wouldn't bet on it.

The blog post exhibits the same flawed logic that many academic reviews do — looking for reasons to reject a work rather than accept it. It comes off cynical and contrarian. It's easy to find issues with an academic paper, but ACL has consistently pushed for a higher standard [1].

Being the author of spaCy, he should know better. I certainly enjoyed the straightforward simplicity of spaCy when I used it in my research (though I haven't touched it in years).

[1]: https://aclrollingreview.org/reviewerguidelines#-task-3-writ...

You've expressed your criticism plainly without addressing anything the post might've done well. I think that's absolutely fine, but why would your criticism apply to the post but not your own comment? Personally I think it's because the standards of peer review don't apply to HN comments, but I don't think they apply to blog posts either.

Some of original blog post really seems to follow Arditi et al. (2024) (https://arxiv.org/abs/2406.11717). Projecting out a direction can only remove behaviour and not add capability so the jump mostly comes from unblocking the looping.

So I do agree it is misleading and trying to present work as groundbreaking.

Just one word - MARKETING

Waiting for

The "The 'Astra can read and write weights to improve Qwen' post is misleading" post is misleading.

I for one think it's pretty cool how far too many people bend the truth or outright lie in tech these days.

There is going to be a reckoning for tech in the near future, because people will remember the lies and it's gonna be hard to beat the history.

It doesn't really matter how accurate the post is, what it shows is what one could assume how you'd end up with a system such as RSI, how you could prune harmful capabilities from models and build a more efficient training system with biases including and not limited to emergent behavior in astra that might have come from something they've never even trained the model on regarless if ML engineers can already do similar things.

None of the previous studies applies because Astra simply did not exist and this is worthwhile NEW knowledge that has been publicized and it is very interesting to see how it can be compared with other models as both openai and anthropic are keeping any and all RSI development completely private. The work is groundbreaking in a way that there will be a lot more experiments of this nature which will contribute to increasing the public knowledge of such systems.

How do you prune out the 'fails to prune out behaviors' behavior? If my model is failing to prune it out, can another model prune out that behavior?

With normal training you're simply training for higher loss for things you don't want while keeping the loss low for things you do. Same would apply here except you would target low eval for behavior you do not want the model to have rather than trying to increase eval.

LOL. Tell me you want to be at the top of HN without telling me.

I don't get it. Because python was used, it's supposed to not be cool? IMHO, working with raw weights with Python is still working with raw weights.

It's exactly what the post says. I was waiting for my excitement to drop for the original research, but I concluded that I was mislead more than this "coverage" than the weight editing research, which was in opinion quite easy to grasp.

Who watches the pruners?

Misleading how exactly when the original post states _exactly_ the same things as outlined in this objection?