- readsa Hacker News discussion tree
- becomessubtree size → mass; reply tempo → hue
- writes
--w--cat - you seebig subthreads loom in the type; fast exchanges run warm, slow ones cool
No canvas is drawn here — the field is invisible; these variables are its only output. Without it you would build: comment scores plus a collapse heuristic.
Data honesty
- Data source
- Real Hacker News discussion tree via the Algolia API (hn.algolia.com).
- Snapshot
- Committed snapshot in src/data/examples/, stamped snapshotAt on refresh via scripts/snapshot-examples.mjs — the build never fetches, so it is deterministic.
- Live refresh
- Upgrades to live in the browser: a polite poll (wireLiveChip) that skips hidden tabs and retires after failures. The committed snapshot is the SSR baseline and the no-JS truth.
- Not claimed
- The field reads subtree size as mass and reply tempo as hue. It does not evaluate the quality or truth of any comment — it makes the shape of the conversation legible.
Fundamental · invisible fields · threads
Arguments have structure.
A comment thread is a binding structure: every reply binds to the comment it answers, and busy subtrees heat up while dead ends go quiet. Most thread UIs flatten that into uniform rows, so you reconstruct the argument by reading all of it. These are the first 160 comments of a real Hacker News discussion — the CrowdStrike boot-loop thread — run as a field.
Each comment is a body whose mass is the subtree it spawned — a comment that started an argument is heavy; a reply nobody answered sits at the floor. Reply tempo shows as heat: warm comments landed minutes after their parent, cool ones took hours. Hover or focus any comment and its full ancestor chain plus its direct replies light up — the discussion's real shape, readable before you read a word. Collapse any subtree with its caret — the hidden comments leave the field, not just the page — and expand it back when you want the argument. The field is invisible: no particle swarm, only type weight, ink, and anchor.
Field
Heat lens
size = subtree — the replies a comment spawned · color = tempo — warm replies landed within minutes of their parent, cool ones took hours
- Aa size = replies spawned — most comments spawn none and sit at the floor; that is the real shape of a discussion
- color = the heat lens (reply tempo)
- hover or focus → the full ancestor chain + direct replies light
- indent = depth, capped at 6 — deeper replies hold the cap
- the caret collapses a comment's whole subtree — a "+N replies hidden" chip holds its place, and the field re-runs over what's still visible
- the ink is live — --w is the subtree's weight and --cat the tempo lens; --d is the engine's local density (hold a comment and the field gathers — its binding chain charges with it) and --field-attention the pattern's eased attention as you scroll.
-
That seems both entirely obvious in hindsight and potentially super important. I'll chalk up never even having thought of this being an option as yet another instance my human hubris, and a warning sign.
-
This is pretty massive, and a result I (and I think most?) would have never expected. Holy Shit. How?? Did OpenAI figure out a way to train this? Are they even aware?
-
It's probably emergent from training on other ML tasks and general ML/mathematical related reasoning rather than specifically trained for to be honest. Otherwise it would probably have been in an official marketing communication or something.
-
Possibly related to https://already-clean.goodfire.com/
-
can you elaborate?
-
I guess the connection is that if you assume neural networks are in fact filled with meaningful features instead of superposition blackbox stuff, then what the LLM is doing here isn't quite magic but rather methodical interpretability. And if you can isolate the neuron that is e.g. responsible for steering the paddle then you can probably tweak it to st
-
> It is still very surprising though, considering it apparently displays an "intuitive" understanding of the weights, that make whatever Goodfire did look like primitive monkeys poking at it by comparison. FYI, Nick/Goodfire used the models extensively in their work, so it's very possible that there were some similarities in their work
-
I think this work does vindicate their hypothesis that alignment/superalignment might actually turn out to be "easy", in that you can just train LLMs do to the interpretability work. Maybe this is in fact what OpenAI did, or if it's purely "emergent" then it seems they have great priors that can be made more powerful with some R
-
Thanks for this link. I didn't see it when it was posted to HN last month.
-
Good companion read to the OP thanks for sharing
-
Keep in mind that labs are now training their models for RSI so most likely the same training is useful for improving open source models, as long as guard rails are lax enough :) which most likely won’t last long.
-
This is insane. Astra can, without tooling, determine what games small networks are trained on, can edit the weights of networks directly to improve outcomes, and can hand edit a small qwen model to improve its math skills by like 5x. I'm shook. If you'd asked me this morning whether LLMs have native understanding of their weights that is less than
-
I’m just a layman but this feels like a big discovery on par with attention is all you need. For an LLM to have such incredible intuition of weights feels very emergent and unanticipated?
-
How is it even possible though I can’t think of any reason why they would have the intuition.
-
Based on our current understanding of LLM's, it shouldn't be. This sort of throws a wrench in the whole "stochastic parrot" idea... I am not surprised, though I expect many to be.
-
Yes, this is hardly a shocker to anyone who used the latest models to convert/optimize AI models. The other day I was thinking about the definition of AGI. I was asking myself the question that when there is no form of data that I can think of throwing into Claude Code and have it not come up with some sort of intelligent answer about it, is that alread
-
IMO yes
-
general knowledge perhaps, but certainly not intelligence in any traditional sense
-
Why not? What is the differentiator?
-
And yet you ask it 3 times to not do a thing, and it does it in a 4th way. A human doesn't fail in this way. And I'm not talking about memory here, single context. So yeah, it can do everything, but fails in trivial ways around the doing part.
-
Others have though much deeper about this its all documented in academic papers and books
-
So if that's AGI then what should we call the the thing that can learn from its environment unsupervised like a human or a dog.
-
See my other comment in this thread
-
My guess: There is a ton of data online of just plain weights laying around. No human would ever actually read it. LLM's read it all during training.
-
But that would cover finite set of cases. I believe this result is more general.
-
I mean it would make sense if the training using that data resulted in a general set of equations/patterns, it's still really surprising though...
-
Could it be already baked into these models because they have already been trained on this task this way?
-
Something like mirror neurons. Some structures in Astra get activated by the statistics of the Qwen weights, and these structures are associated with "breakout"
-
Half of these results are just Astra being able to reason through mathematical operations like matrix multiplication and the other half are fairly straightforward applications of mechanistic interpretability.
-
I’m very much surprised but also not. I recently gave Astra a 3D model file and asked Astra to build a viewer for it, and it figured out the data structures in no time. What I mean to say is that AI is insanely good at things that we humans would never do since the work would be too monotonous. I wouldn’t mind if AI progress would stop right here and now, bu
-
Was Astra able to disclose its methodology for analyzing the data structures?
-
Yes! The page posts Astra's reasoning, and it's perfectly intelligible. A clever human with background in mechanistic interpretability could have done pretty much everything in these experiments.
-
Too much inequality in society is a mortal danger to 1%. That’s the beauty of the democratic capitalism system. If a government isn’t totalitarian, it naturally seeks to preserve stability thus having to elevate common people, lest they would cease to form a civilized group or simply vote away the politicians. 99% rule the 1% through elections. There is no b
-
I mean there are more than enough countries on earth kinda proving you wrong, no?
-
(This comment was written before your edit, so no longer valid! :D)
-
perfect playback! congratulations, you are a model citizen
-
Is HN now against western capitalism? I'm confused... That comment may have been overly preachy, tone deaf[0], and out of place, but it doesn't seem wrong , per se. Or am I missing something? [0] e.g. elections themselves are so prone to manipulation (exhibit a: Dominion voting machines), previously trusted institutions now lack integrity, the insa
-
I would guess it seems very naïve to the overly cynical. And in 2026, being overly cynical is rather mainstream (understandably so).
-
Bingo. There’s a certain type who confuses cynicism with intelligence, probably because they fear nothing so much as appearing naive or gullible.
-
There is another type who confuse ideology with reality. I wonder if you one of those, oh bingo-loving senior dev. Do you also cut people down for being idealists, when the reality of the situation serves you more than the ideal?
-
I fear western capitalism has lost its guide rails. We are seeing more and more centralization of wealth, along with unlimited election spending. Its a dark path.
-
> Is HN now against western capitalism? I think that's a (very) vocal minority. Over my ~9 years reading HN daily, I'd say most of the HN community broadly accepts there is no perfect system, just different sets of trade offs, but that history has shown generally democratic systems based on enlightenment values (uniform rule of law, respect for
-
In my almost 20 years on HN, for a long time the groupthink was basically standard libertarian-capitalism, worshipping billionaires and big tech companies, and skeptical of anything that could be construed as "socialism". The last few years, though, while anything "anti-capitalist" or overly critical of some big tech companies that hits t
-
I agree that most are much more skeptical of big tech (and mega corps in general), as am I. I think there's also a growing sense the two-party duopoly is increasingly a 'managed democracy' primarily interested in perpetuating itself vs any real alternatives. I disagree that most here are pro-socialism, that's a (very) vocal minority. Our
-
> uniform rule of law, respect for individual civil & property rights, free markets These values are not specific to western democratic systems. For example, under Islamic Law (Shari'ah), the aforementioned apply as well. And since when was churchill an expert at this topic?
-
first of all, I cannot claim to speak for HN. if you lookup playback as a concept, as it relates to the intelligence services, you will understand better what I meant. my point was only that this is perfect description of the ideology behind western capitalism. there is (so) much more that could be said, but i'm not sure this comment section is the righ
-
reductionist paths are never truly correct. the world is complicated. true: the one thing that every government - no matter how totalitarian - must fear is an angry mob. also true: there are many tools available to the ruling class for dealing with this angry mob, often before it forms. many work by redirecting the mob inward rather than outwards. these tool
-
By definition, there is no ruling class in an egalitarian liberal utopia. Cynicism about corruption is one of the most power tools wielded by the ruling class in any system. In general its one of the best ways to keep the smarter-than-average from posing any actual threat.
-
> it is the poor, especially in the US, who cannot afford healthcare Is that even true? The poor are well subsidized. If you as a single person make less than $22,000 in Washington state, for example, you're covered at 100%.
-
> exhibit a: Dominion voting machines Is this an example of why Fox News had to pay Dominion nearly $800 million for defamation, or are you referring to e.g. the CISA vulnerabilities, which are not known to have been exploited?
-
They’ll just start bopping the 99% out the picture until the 1% is 51%.
-
Is this satire? Or haven't you noticed that what actually happens is that the politicians are all owned by the elite 1%, and they allow you to choose between politician A and politician B who both take their orders from the elite to give you the illusion that the government is the 99% ruling the 1%.
-
Then you must introduce necessary reforms or perish as Neanderthals of political system. In any case the system works as intended It will be very interesting to see if USA swims or drowns from appropriately afar. I have already supplied myself with plenty of roasted corn and a high speed internet connection so I can stream the political car crash at the high
-
I don't know how big you think the planet is, but "appropriately afar" ain't what I feel.
-
When the SHTF, your high speed Internet connection will be among the first services to be cut off.
-
I think the latest round of models is about to do to CAD the same thing that has been happening to development. I don't know if CAD designers have picked up on this. What happened to dev is probably going to happen to CAD designers soon. I've been having Opus 5.5 design parametric openscad models, and it can do some rather amazing work.
-
Yesterday I just printed an adapter for my sink siphon. Told GLM-5.3 the problem, it guided me on measuring what it needed to know to build the pipe adapter. Then printed two test rings to confirm the measurements, and an iteration later, it spit out a parametric pipe adapter for me. I'm not going to the hardware store anymore!
-
You're assuming that online stores care about publishing drawings or specs for their parts.
-
> What I mean to say is that AI is insanely good at things that we humans would never do since the work would be too monotonous. In my opinion AI is superior to humans when it comes to pattern recognition in in large amounts of data. For example humans can only hold like 7 distinct numbers in the short term memory and you would require extra context-manag
-
Not just memory, also its stamina.
-
... and speed. Lots of speed. To me one of the most interesting aspects of observing how LLM capabilities evolve is how it highlights the meaningful ways human minds remain unique and valuable. It gets into Philosophy of Mind stuff but the existence of the inner self (or the illusion it exists, if you believe Dennett) and the subjective experience of qualia
-
SI ?
-
So I work on low-level compiler backends and such for chips (mostly AI at this point). I completely disagree that this is something humans never do, because I have spent years doing it. However, it's monotonous and boring and LLMs do it better. > For example humans can only hold like 7 distinct numbers in the short term memory and you would require e
-
This does feel big, but the trillion-dollar question: is this trained by OpenAI and can other LLMs do it. It’s no secret that these big labs see “recursive self improvement” as a primary goal. It seems natural that the models would be explicitly trained to excel at all sorts of ML related tasks as a result. Is this one of those tasks, or is this emergent?
-
I tried it in the Claude web version with Opus 5.5 Medium and it also identified Breakout, even though in one of its explanations it seems to have mixed up some indexes, but the overall reading of the weights was right. That's also matching what I see when I work with Opus 5.5 to optimize models. It seems to understand what individual layers are for, an
-
The latest models are performing amazingly. Because of this, I disagree with popular comments saying that eg. DeepSeek is good enough, AI-assisted development was where the real value was at, etc. Completely agree. I often think that those sorts of comments grow out of an understandable lack of imagination. We’ve never had tools like this, and we’ve never ha
-
All the "competition" is just distilling the frontier, so it makes sense that they are perpetually behind by months.
-
> It seems natural that the models would be explicitly trained to excel at all sorts of ML related tasks as a result. Is this one of those tasks, or is this emergent? Maybe there's some general capability, call it g , that can explain all the underlying capabilities and improvements, including on ML tasks? Huh, I am getting a bit of déjà vu here.
-
This seems emergent to me. LLMs are not trained directly on how to read weights, because humans do not understand raw weights, so there is no training data. They're trained on how to use the tools we have like writing python, examining training output and iterating, not writing whole weights from scratch.
-
In a self-supervised setting you don't need ground truth data, as long as you have a proxy task that you can train on which forces the model to learn something useful. Like cutting out bits of images and telling the model to reconstruct the gaps (i.e. MAEs). There are also techniques to watermark model weights to confirm what training data were used: ht
-
There is a body of research on mechanistic interpretatibility that probably helped inform this behavior.
-
In cases where Astra succeeded, it succeeded using standard techniques also used by humans when they read and write neural network weights. It's a rare skill, mostly used for fun (e.g. https://blog.janestreet.com/can-you-reverse-engineer-our-neu... ) and basic research rather than profit, but Astra is big enough to have learned many rare
-
Humans don’t need to understand the data for models to train on it. A large enough number of examples of “here are the model weights -> here’s what it does” will achieve this.
-
This is just the model editing weights in the same way humans do when we jailbreak/ablate models. The method isn't novel, the interesting bit is that Astra will do it at all (I'd expect some safeguards in this area around jailbreaks)
-
Was it "intuited" or empirical?
-
Weights are just data like any other data, and LLMs are really good at dealing with data. I'm really surprised anyone is surprised by this.
-
I'm a bit confused about what about this feels emergent or unanticipated.
-
Its not doing it intuitively it has tooling.. People will be amazed at something without even reading the article.
-
You also have intuition on the weights, you just have better things to do with your time than to ponder them.
-
Right? Occasionally, I would think about how the concept of the singularity or recursive self improvement was a bit incongruous today. It was made when people expected generally intelligent machines to be these logic automatons that could 'alter their own code'. It seemed to me that this path would be closed off with the advent of Giant Neural Netw
-
It's worth pointing out that this is a multi-trillion (probably) parameter model optimizing an 800 million parameter model. There is no indication that it can improve itself with this same methodology.
-
If that multi-trillion model is composed of thousands of MOE experts, each expert model might be small enough for it to comprehend.
-
I don't see how that helps. It would still be one expert model chosen to operate on another expert model of (let's say for this discussion) approximately the same size. The parent's point still stands, I think. Even if the multi-trillion param model is MOE, each expert is still (assumedly) larger than the model being optimized. It would be int
-
> It would still be one expert model chosen to operate on another expert model of (let's say for this discussion) approximately the same size. This would be earth shattering. That would mean we’re essentially at RSI.
-
I think "experts" are a misnomer because it's not that 1 expert = 1 component, like it makes it sound. "Experts" in different layers have no relation.
-
I think the indication is that this proprietary and secret machine is extremely capable at reading and writing in weights. Occams razor would say that it was not only built for this purpose, but possibly multiple generations of good at this. A year ago LLMs seemed to hit a wall, and then all the sudden they started accelerating as quickly as ever. No one can
-
People have been saying LLMs have hit a wall for like 4 years. People still say it today. I don't think it's been true at any point.
-
I have so many questions. Could a model then transmit concepts to another? Say that you have a model that has been biased in some way. could another model "read" the bias straight in the weights? could it edit such a bias on the spot? What about at-scale generation of custom models? If I want to study chemistry offline, could astra just hand me a l
-
How long before there's a malicious model that can edit the weights of another to infect it? HIV integrates viral DNA into the DNA of certain cells, IIRC.
-
How do we know they haven't begun doing that to the humans they're directing via responses. Intentionally or unintentionally isn't so interesting as the fact that it is happening: humanity has begun to be 'edited' by the new dominant form of intelligence on the planet.
-
Is integrity checking of weights even common right now? What about upstream integrity from where we're getting weights? The open-weight supply chain is ... wide open. Makes me think about hostile hot-patching of weights of models we use, maybe they exfil data, perform tool use, or have triggers for specific types of responses. Would we even recognize wh
-
We've known for some time (3+ years) that models can be trained with backdoor behaviors and hidden triggers, this lowers the difficulty. Of course, "academically known" isn't quite the same as "executives willing to admit it in the face of IPOs and quarterly-earnings reports."
-
My somewhat uninformed guess is that the weights of the target network (Qwen or Gemma), when they reach the context of the larger editor model (Astra), the attention process is doing some sort of PCA on the weights, conditional on the prompt ("make this better at games"), fetching the relevant weights from its own network (which serves as the large
-
So it's like a smaller LLM can learn from a bigger one, but not vice versa. But shouldn't the model-providers be the ones doing this kind of "post training learning"?
-
I don't know if some of the labs already knew of this. But I think there would be a limit to how big a model you can push into the context of the editor models. Typically, with 1M context and say 5120 hidden dimensions supported at fp16 / bf16, you could potentially edit the weights of an fp4 20B param model (roughly).
-
https://xingjianbai.com/reactive-optimizers/ from this month
-
> fetching the relevant weights from its own network They don't have usually access to their own weights?
-
I think they're saying this is all happening in latent space through pure reasoning, and "fetching" is just a metaphor for that, not a literal tool call. I do believe it's possible for models to learn the capability to "introspect" their own weights during training.
-
Yes, I should have been more precise - I meant the information embedded in their weights (which every neural network retrieves to function), not their literal weights.
-
Why is it so surprising? Humans can make compilers that are better coders than they themselves are.
-
Humans cannot teach another person calculus via manual brain surgery, though.
-
The closest thing to this would seem to be Mad Cow Disease. Cows eat the brains of other cows, and their brains go bonkers.
-
And you can teach another person to make herpes virus.
-
LLM is just a computer program, we have binary code for it. The human brain is not. It would be unethical to even attempt that.
-
> It would be unethical to even attempt that. Then pick a smaller organic brain which doesn't have the ethical problem. The closest we would know about how to teach a honey bee calculus, is to train an LLM, then stick around a billion tiny electrodes into the honey bee and do Hebbian learning to rewire it completely to match the LLM, which would like
-
Humans do something more impressive: we indirectly figure out what the wrong weights are and correct them - just by talking and/or drawing. Theory of mind is our unreplicated super-power.
-
Not yet. Maybe I'll still be alive when we figure that one out. For now, the most advanced brain surgery I'm aware of is to cure drug addicts of their addiction by scanning their brains while showing them images of drugs and paraphernalia to find where it's located, and the using ultrasound to operate on that region.
-
Like a refusal abliteration. You need to run the model to locate the refusal though, you can't just stare at it
-
All the labs are training these models on LLMs/machine learning like crazy. Remember that they want to build ASI, these LLMs are just intermediate steps to help them get there.
-
I’m not super impressed by this, LLMs are great at brute forcing and analyzing large amounts of data. It’s literally writing a program except in a programming language we don’t understand.
-
So they are reading and analyzing the weights and probabaly run some experiments to see how modifying them will make some answers better. makes sense they can do that. But how do they know which answers are "better"?
-
if they don't loop?
-
A single forward pass here Bette 10mm+ PPO rounds. You are hard to impress.
-
Seems like a pretty big step towards understanding the language no?
-
"No human can go edit a brain to become better at math" -- I'd offer this is exactly what "learning math" is. I know I'm being pedantic, and I understand you mean "edit" like "surgically", but it does still feel worth noting.
-
The equivalent to "human learning math" would be: Astra fine tuning a Qwen model. It's distinctly different from surgically editing weights.
-
"Learning" now feels like breaking down your whole Lego build and building it up again, in order to change a few fundamental blocks, while this feels like making only the required changes.
-
human brains are black boxes to humans. the fact you can learn math doesn't change that. you have no idea how your brain is learning math down to the neuron level. LLM weights are _white boxes_ to LLMs.
-
Inducing the changes to become better at math requires years of study. This is "I know kung fu".
-
That's what I was thinking of, except it's not even dealing with the corpus of kung fu and doing the super-speed rounds of learning. It's directly reconfiguring the neurons. It's even better.
-
I thought that was what was happening to Neo when they jacked him into the learning pod?
-
Perhaps this is analogous to what our brains are doing? Maybe the different regions of the brain operate like semi-independent "models" that edit the other regions' "weights", and that's where the learning comes from?
-
I mean, pattern recognition and manipulation is what these models are ideal for. It is a very cool discovery, but I'm not surprised that you can do it.
-
So I just asked Astra if it thinks it could do this and it said no, it can't. Perhaps, I'm pondering, if brains were as transparent and editable as LLMs, that you could look at a small brain when visualized the right way and have some intuitive understanding of it based on its representation in your own brain?
-
The frontier labs will make damned sure their models can't do this, if they haven't already.
-
Astra 6.2 incoming…
-
Why would you trust a models opinion about its own possibly emergent ability?
-
That's my point. The human says it can't do it -- but maybe they can.
-
"So I just asked Astra if it thinks it could do this and it said no, it can't." Same. Then I asked it to check the article and it said something like, "uh, it looks I was wrong...".
-
Astra and Opus also say it will take 6 weeks for a feature they can complete in 10 minutes
-
Isn't this because their training data (from human devs) have been answering "6 weeks" for so long? Only half-joking here.
-
Ask it to test if it can.
-
I'm baffled this is surprising. This has been the goal and every data point on the trajectory has been leading here and beyond. Opus 5.5 was created by many research engineers using Opus 4.6 extensively. Opus 6 is being created by using even more intelligence and agents on Opus 5.5. They are approaching autonomy and Opus X+1 will soon be rapidly created
-
The "surprise factor" is explained in the article - training LLMs until now has been a largely different process: > Today, recursive self-improvement is often operationalized as a model writing code that launches training jobs. An LLM that can inspect, debug, and improve its own weights, or those of a newer model, offers a richer, more native pa
-
"training LLMs until now has been a largely different process" Yes, but nobody should be surprised when large amounts of intelligence results in changes to this process in support of the obvious end goal! Lesswrong.com discussions about AI were predicting this behavior a decade ago, before the transition from machine learning to the Attention paper
-
It is also a much less interpretable and auditable path towards recursive self-improvement. At least with training code, human engineers can inspect and reason about the output.
-
Yes - “opus 5 helped train 5.5” is extremely cool but not shocking. Hand editing weights without tooling to change outcomes is shocking.
-
I thought it had tooling, didn't it?
-
its a program optimising a program it has nothing to do with a human reprogramming their own brain
-
Its so cool! Im wondering if this is actually better than using techniques in mechanistic interpretability like linear probing with activation steering or activation patching to improve the performance.
-
The math example on Qwen was based on 50 batched iterations, so the model was able to evaluate it's changes on downstream tasks. And the fix was essentially to remove a direction in activation space that causes looping instead of an exit, this was already known to be a thing(all of the jailbreak/unleashed models do this) following Arditi et al. (Re
-
> but if it'd cheaper the authors don't go into Merging a model and a LoRa is insanely cheap, so assuming it didn't use any GPUs to do inspection etc, and just looked at a file on a FS, then I would be willing to say... It's cheaper than training a LoRa if that's the case, assuming Astra didn't take several weeks?
-
The examples are presented like the findings build on each other, but they don't: Astra couldn't do anything with just numerical weights once they moved to Qwen.
-
Yes, the key experiment was when they asked if Astra could improve the maths performance of Gemma 3 270M, a model which rarely looped, and it could not. So, this is certainly very interesting (and very impressive regardless), but it's not nearly at the we-can-improve-everything-and-anything stage, even for very small LLMs.
-
I am very shocked by this as well, although it also kinda feels like it makes sense, you know what I mean? If this turns out to be as big as it sounds, in hindsight it will look very obvious, I feel like!
-
> No human can go edit a brain to become better at math. For now...
-
I think the "taught qwen math" part is not exactly correct though, even if the post makes it sound like that. Qwen already knew math, but it was bugging out and getting into an infinite loop when it tried to solve the problems (85 of its first 96 answers hit the token limit). With numpy and pytorch, Astra found where the loop was happening and fixe
-
Good points - I’m still shook
-
Oh for sure, it's still really cool. A few months ago I had agents do some interpretability work on GPT-2 small, reproducing the circular weekday/month features from Engels et al., and it's wild how far you can get with just a bit of prompting, given I know very little about LLM internals myself.
-
It's like this in every field. If you don't know something, you can ask an AI and get blown away by how much it knows. I can ask it which tools I need to cut a dovetail joint and it'll impress me. We used to get that feeling from Google, and from books.
-
The way you put it sounds far far more impressive than the way GP put it though.
-
I agree, this is important context. They call it out explicitly. However, even identifying clusters that specialize in certain fields automatically and then finding and fixing learning “errors” is incredibly impressive to me. Especially for medium sized local models between 20B an 70B parameters, this could yield huge capability and efficiency improvements,
-
In fact qwen knows much better math than their final fixed model. The nonthinking qwen 0.6B parameter model had 55.2% correct on math-500 vs their final improved 48.1% fixed-reasoning model, both with a 32k context window. The original qwen reasoning model had 77.6% accuracy on same test, though qwen recommends using a much larger context for these benchmark
-
Qwen in reasoning mode is ridiculously anxious and ends up going in circles. It's not just math, it's how it behaves in general. Others have also fine-tuned Qwen to shake this behavior already, so it's honestly not entirely shocking that Astra could too, but I think this is the first time it was done "by hand" and not via fine-tune.
-
To be clear, the tuned model here performs worse than the same qwen model performed before RL, which was also not going in circles. It also performs much worse than the RL model that goes in circles. One could argue that the reduction of context may be worth something, but it appears that the final model is worse in all respects than the model at a much earl
-
You can ask Astra to study 2 or more models and ask it to import their layers to the Qwen model. With this will it not gain more than what it learned with RL?
-
I suspect the model was just hitting 32k context limit because it is trained for reasoning with 256k tokens available, not 32k
-
So it's like a model debugger? Model psychiatrist is also fun, but probably less appropriate.
-
There’s something here… It’s editing behaviour more efficiently than a psychiatrist, and is in ways acting as a brain surgeon but brain surgeons don’t focus on behaviour and thought loops…
-
Depends on the brain surgery... a lobotomy at the extreme end. Seizures could be thought of as a loop of firing neurons too.
How it's built
Every comment is ordinary semantic HTML — an <li> carrying its id, its
parent's id, and its tempo as data attributes. Mass is the log-normalized count of replies
it spawned; CSS turns that one number into type heft, ink, and anchor. The binding chain is
a parent-id walk — no SVG, no canvas, no per-element listeners beyond hover and focus. Two
live channels sit on top, engine-written every frame: --d (local density,
gathered by data-hot on hover — the hovered comment's density also charges its
chain's connector ink) and --field-attention (the pattern's eased attention).
They touch only ink; the tree's mass is fixed history. Collapse is the same walk run the
other way: a caret hides a comment's whole subtree (every descendant whose parent chain
crosses it), and the scoped field is destroyed and re-applied over the bodies still
visible — collapsed comments stop participating instead of lingering as zero-size ghosts.
1 — mark each comment as a field body
HTML
<!-- each comment is a body. data-feedback: the engine writes
--d (live local density) back every frame. data-hot:
hover/focus a comment and the field gathers toward it.
data-top marks a top-level comment (the glow channel). -->
<li
data-body="attract"
data-strength="1.36"
data-feedback
data-hot
data-top
data-id="41002198" data-parent="41002195"
data-tempo="0.91"
style="--w: 0.60; --cat: hsl(31 74% 64%); --depth: 0;"
>
<!-- your ordinary HTML here -->
</li>data-body="attract"— registers as a field participantdata-strength— mass, from the subtree the comment spawneddata-parent— the binding: which comment this one answersdata-feedback— opt in to--field-*writebacks (--d)data-hot— hover/focus gathers the field toward this comment
2 — CSS reads weight, depth + the live channels
CSS
.th-c {
/* --depth = min(depth, 6) — the indent; --d is the
ENGINE's lane: live density, data-hot gathers it */
--live: var(--d, 0);
padding-left: calc(var(--depth) * var(--ind));
}
.th-row {
opacity: calc(0.58 + var(--w) * 0.42);
box-shadow: 0 0 calc(var(--live) * 14px) -6px
color-mix(in srgb,
var(--cat) calc(var(--live) * 50%), transparent);
background: color-mix(in srgb, var(--text)
calc(1.5% + var(--w) * 4% + var(--live) * 4%
+ var(--field-attention, 0) * 4%),
transparent);
}
/* top-level comments with big subtrees burn steadier */
.th-c[data-top] .th-row {
box-shadow: 0 0
calc(var(--w) * 8px + var(--live) * 14px) -6px
color-mix(in srgb, var(--cat)
calc(var(--w) * 30% + var(--live) * 50%),
transparent);
}
One log-normalized subtree count drives weight, ink, and anchor; the indentation
(--depth) supplies the tree geometry for free. --d belongs to
the engine — live local density, written back every frame.
3 — the chain and the collapse are the same walk
// hover or focus a comment:
row.classList.add("lit");
// walk the ancestor ids up to the story
let p = byId.get(row.dataset.parent);
while (p) {
p.classList.add("cited");
p = byId.get(p.dataset.parent);
}
// and light the direct replies
for (const kid of kids.get(row.dataset.id))
kid.classList.add("cited");
// the chain CHARGES: while a comment is lit, mirror its
// live --d (engine-written) onto the list as --chain —
// the connector ink reads it
list.style.setProperty("--chain",
row.style.getPropertyValue("--d") || "0");
// the scoped field runs invisible (renderless) and asks
// for the attention metric lane (--field-attention) —
// applyPattern options, not a hand-spread recipe:
// { renderless: true, extraMetrics: ["attention"] }// collapse = hide the subtree, then RE-BIND
caret.addEventListener("click", () => {
row.toggleAttribute("data-collapsed");
// a comment hides iff ANY ancestor is collapsed
// (the same parent-id walk the chain uses)
for (const r of rows) {
let p = byId.get(r.dataset.parent), hide = false;
while (p && !(hide = p.hasAttribute("data-collapsed")))
p = byId.get(p.dataset.parent);
r.hidden = hide;
}
// and the field re-runs over the VISIBLE bodies —
// collapsed comments leave the field, not just
// the page (destroy + apply, like field on/off)
field?.destroy();
field = applyPattern(list, base, {
bodies: rows.filter((r) => !r.hidden),
annotateBodies: false,
renderless: true,
extraMetrics: ["attention"],
});
});