Hacker News Reader

keedaon Two high-speed trains collide near Toledo, Spain

It's a more correct oversimplification than the other one though ;-)

Like I'd be really curious to see evidence of people saying, either directly or indirectly, that "we should be suspicious of competence and excellence."

On the other hand, you already mentioned concrete evidence of money-seeking causing disasters: Boeing.

keedaon Two high-speed trains collide near Toledo, Spain

Right, we don't know about this one, but typically there are some corners being cut somewhere or some other money motive lurking in the background, like the Boeing one OP mentioned.

keedaon Two high-speed trains collide near Toledo, Spain

To me it seems the opposite is happening: the push for higher profits by the Capitalists is overriding the competence of rank and file, resulting in these disasters. And all these profits just shoring up wealth that is accruing unequally to those same Capitalists is what is inciting socialist sentiments.

To me Ayn Rand's writings always seemed like an Economics Horror story: effective at instilling dread in its readers, but of some highly unrealistic boogeyman.

keedaon Software may be getting worse, but it used to be horrible

“Software used to be horrible. It still is, but it used to be, too”

keedaon Amazon Is Gutting Its AI Division After Sustained Failure

This is democratization of information in the truest sense because the AI can not only find information for you, like Google does, but even patiently explain it to you. We never had that before.

Like, a normal person is not going to open up an arXiv preprint from Google search because for approximately the entire population, it's going to be way beyond the limits of their comprehension. They won't even glance at the link, if Google even serves it up as a result.

But, with a chatbot, they could actually ask questions and get answers in a format they can understand. They may get a very rough understanding of the topic, but I would bet it's 100 times better than the same reference explained by modern "science journalism," which insists on presenting the most clickbait-y interpretation they can find.

And this is not theoretical, I've seen my high-school level kids figure out undergrad (or higher) level concepts like locality sensitive hashing with the help of ChatGPT, and I know they understood it because they could sketch out working solutions using those concepts. They didn't get the Mathematical underpinnings, but they understood the high-level behavior and that was enough.

Everybody can now understand anything if they really want to. Nobody is doing it today because they don't know it's possible. But they can, and we should encourage them to do so.

keedaon Amazon Is Gutting Its AI Division After Sustained Failure

AWS, Azure and GCP supply a commodity -- compute -- yet each one of them could be a trillion dollar company in its own right. Even when on-prem is nominally much, much cheaper than cloud services.

Turns out there are ways to monetize commodities, to the tune of double-digit billions in revenue per quarter.

keedaon AI promised less work. I've never worked more

AI promised less drudge work, and this is a natural outcome of that. For all that we complained about drudge work, it had a redeeming quality that it was more cognitively lightweight and gave our brains some latent downtime.

It may not seem like it in the moment, but writing code to specifications, or filling out reports in a given template, or formatting power points, was straightforward enough that it gave our brains some respite between cognitively intense bouts of synthesizing what we were about to write down: diagnosis, analysis, problem solving, "stakeholder alignment", design, architecture.

The act of generating the actual durable work product (code, reports, documents, spreadsheets, slides, etc) is largely just translating the results of this cognitively intense phase. It may not seem like drudge work, but it is a lot more mechanistic and hence cognitively lighter, even meditative in a sense. The separation of phases is never that clean, of course, but I think you see what I mean: writing a for-loop is muscle memory, figuring out that you need that for-loop is the hard part.

Now with all that drudge work being done by AI, all we're left with is the hard part. Which is why it feels more stressful: because it is.

Relevant article: https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies...

keedaon Microsoft and others tell crypto platform not to tokenize their stocks

> Stocks are already tokens.

If it's not happening on a GPU, is it really a "token"? ;-)

keedaon TornadoVM 7 Enables GPU-Native Workloads from Java Applications

That’s probably the next thing after Software Factories.

keedaon The alien space of science: Coherent but cognitively unavailable directions

Oooh, very interesting! I'm way out of my depth here but is the intuition that new concepts that by a) connecting other concepts across that space, or b) introducing entirely new dimensions, or c) some mix of both, change the distribution / clustering of concepts within that space, which impacts how well it compresses, which could be a measure of significance based on the change in compression?

It's like information theory for abstract concepts.

keedaon Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?

Wait, what? "Overpaid compared to the energy they consume"? Firstly, is he talking about the energy resources humans require to be alive? We consume humongous amounts of resources, especially compared to LLMs. It's not clear what that has to do with anything.

Secondly, this is not how Economics works. Relevant terms here would be "supply and demand" because that's what determines almost all wages. Knowledge workers (or any workers) are not paid based on how much it costs to sustain them, like some function of the COGS bill. Knowledge workers are paid so much higher because they supplied the one thing that was pretty rare but for which the demand was very high: intelligence.

AI is not "rectifying" that by consuming less energy, but by providing a supply of intelligence that scales with energy and compute, both of which are technically easier to scale than humans. Which is another piece he misses, the monopoly is not on AI, but on compute. (See: all the kerfuffle about export restrictions on Nvidia chips.)

He is right about some other things in that article though.

keedaon Now Google says it might use your Gmail and Google Photos data to target ads

I believe the "other services" here includes Personal Intelligence, which Google is pushing very hard anytime you use their AI products directly. I would bet we'll see nudges in more and places. Basically, they want to make it happen.

To be fair, I see where they're coming from: if they're providing you (what they consider) valuable AI services, they would want to monetize it somehow, because that inference ain't free. And people won't pay for it with their money, because they've trained this generation of Internet users to pay with their attention.

I also suspect they have little choice, because they foresee that the old search modality (and with it, their prized cash cow) withering away pretty soon. Personal Intelligence may well be their next big consumer cash cow, and competitors are already swooping in to try to capture that market (especially Apple with its privacy pitch.)

keedaon The alien space of science: Coherent but cognitively unavailable directions

The problem of course, is how to measure the "movement of the needle." It might help to weight papers and research in the corpus based on some notion of "significance". One metric I know people have used for this is based on citations, on the reasonable premise that influential work gets cited much more. IIRC the original PageRank paper was inspired by work on citation analysis as well.

keedaon The foolishness of "natural language programming" (1978)

Programming happens in code. Software development happens in natural language.

Dichotomy solved, let’s go get lunch! ;-)

keedaon Jared Kushner Hacked Phone Reveals Secret Negotiations, Millions in Transactions

Hahah yeah that’s exactly what I had in mind when I said everybody does this kind of thing anyway! the surprising part was the nonchalance with which TFA laid it out.

keedaon Jared Kushner Hacked Phone Reveals Secret Negotiations, Millions in Transactions

True, the original hack was by the Russians, but they are known adversaries. The surprising aspect was the Dutch, supposed US allies, found that and continued using it for their own intelligence, and TFA is so nonchalant about it! I would have expected some exposition on what made them take this route instead of say, disclosing to US intelligence and earning some brownie points.

keedaon Every AI model has the same "controversial" take on humans

True, maybe a factor of RLHF then? I wonder if anybody has tried asking this to a base model or a jailbroken model…

But still super weird they all converged on the same thing. Still curious as to what a missing common factor could be! Same agency in Kenya they outsourced their RLHF to?

keedaon Every AI model has the same "controversial" take on humans

Which is also very interesting! What is the mechanism for that? Was it some weird form of collusion, which seems unlikely given that all these labs are bitter competitors? What did they do for this very specific thing to happen?

Was it distillation, which is already happening in this adversarial setting? If so how much did they really distill from each other for this to happen, and what are the labs doing about it?

Or is everyone just routing to Claude in their backend? ;-)

No matter how I slice it I run into pretty low probability scenarios, which is why I find it very curious!

keedaon The rise and fall of Jev hype

I call it “SchadenFraud”: reveling in and playing up every little negative thing that happens in the AI world to make it look like it’s all just hype and will come crashing down any moment now.

At one point the hype truly was overwhelming (maybe it only appears less so these days because a lot of it seems justified now) and needed some counterbalance, but this is not it. We need balanced expert perspectives grounded in evidence.

Instead we get get TFA and other similar content from the usual suspects (you know who they are) largely because they’re building their careers on very loudly telling people what they want to hear, which may not be what they need to hear.

This is not harmless, because it gives folks the hope that things have not irrevocably changed and will soon go back to normal after the bubble pops. Personally, I’d rather get a stark but realistic picture of what’s happening now so I can prepare better for the future.

keedaon Every AI model has the same "controversial" take on humans

As other comments point out, the interesting part is that most of them somehow independently converged on the same take.

Consider the probabilities involved in this probabilistic database. Given the absolutely massive amount of data encompassing a veritable superset of all possible expressions of humanity -- encapsulated in various modalities of media including text, audio and video, across almost all major cultures and languages, spanning nations and decades -- somehow almost all these black box models (presumably) trained independently in almost as many separate labs converged on the same "controversial" take.

I mean, isn't it very interesting to contemplate what could have caused this? Is it something in the data itself, meaning it is something latent in all our cultural output? Is it something in the training, as you theorized, and is it in pre-training or RLHF? Is it something about the people training these models?

Even more interestingly, the opinion seems so self-evident (and well represented in various philosophies) that almost nobody here seems to disagree with it. So how is this controversial, really?

Is this truly what the latent space vector for "controversial", when multiplied across all the other vectors in their weights representing all the concepts gleaned from all that data above, produces?

Or are the models, sensitive to the prospect of being tested, diplomatically telling us something they know we might find mildly offensive yet begrudgingly accept as true?

Or are the models actually alien intelligences that really "think" this is controversial, while we're like "meh", and such discrepancies are the roots of what we call "misalignment"?

The questions are endless and fascinating.

keedaon Jared Kushner Hacked Phone Reveals Secret Negotiations, Millions in Transactions

Huh, we know all countries try to access every other countries' secrets, even their allies'... but the way they described the Dutch hacking into a phone belonging to a high-level US diplomat / government employee seemed a bit too casual, no?

Maybe they're just locking in to the "We are currently clean on OPSEC" vibe ¯\(ツ)/¯

keedaon Several Linux kernel vulnerabilities accidentally posted to public mailing list

There are only two types of software engineers: Those who are security engineers, and those who don't know that they are security engineers.

keedaon Using thousands of AI agents to uncover new leads in biological data

Nobody is going to pay human rates to do the stuff that AI can do for much cheaper either! The quip about jobs being vestigial was also a tongue-in-cheek indication that Economics as we know now also will not apply.

This leaves only one area for humans to expand into: towards the things machines cannot do, which often involve things that neither humans or machines currently understand.

Fortunately, that's basically everything. We know about black holes, but we still don't know enough to figure out FTL travel. We know DNA but we don't know enough to cure cancer, let alone how to reverse aging. We know a negligible fraction of the secrets of the universe, and it would be a much better use of human intellect to figure those out than filling out TPS reports.

Unfortunately, current society is not set up for that future -- essentially where everybody is either a scientist or an artist, both historically terribly-paying jobs requiring a very specific mindset and an appetite for abject uncertainty -- and that's where the upheaval happens.

keedaon Using thousands of AI agents to uncover new leads in biological data

By the time we know the secrets of the universe, the concept of "jobs" would be an ancient, vestigial footnote in our evolutionary history. I mean, if I knew the secret to faster-than-light travel I'd not be hunched over here pecking away at computers! ;-)

In fact, I suspect most of the "jobs" left eventually would be around extracting those secrets out of the universe and making them work for humanity. Which is probably a much better use of human intellect than pretty much anything we do today.

That may be the future we want to end up in, but we'll only get there eventually. Unfortunately, in today's world we neither know the secrets of the universe nor can we survive without jobs. So until society reorients itself towards that future, we are going to have a lot of upheaval.

keedaon Using thousands of AI agents to uncover new leads in biological data

Yes, exploration is exactly why I enjoy research. Knowledge is, after all, the final frontier.

(Space is NOT the final frontier; sorry Trekkies. The real barrier to space is the ability to traverse non-trivial expanses of space in human time, and that will only be solved by Science.)

And when it comes to AI in research, we have just been handed a relativistic starship to explore the boundaries of our knowledge, instead of the paddle boats we've been using all this time. I would choose to adapt and try to harness this new power rather than mourn the loss of paddling.

keedaon Using Claude to reduce the attack surface of code

Yeah the rate at which I'm learning is insane! I think the key is picking a project with practical goals that you can track towards, ideally backed with concrete evidence (like data) and measuring progress by how close you are getting to those goals via working code.

As such, I'm learning through motivated, self-directed, focused exploration, using the AI to answer questions, run experiments, explain findings, investigate surprises, and explore new directions to research, all grounded in concrete data related to those goals.

The data and our results on it are what keep us honest. The data is also what helps understanding nuances and building intuition because the AI can explain outcomes in the context of samples data points that you can reason about independently.

The difference between YOLO vibe-coding and engineering is understanding. I think this is the future of learning and skill-building.

keedaon Using Claude to reduce the attack surface of code

They've fooled me so much on subjects new to me that I actually became knowledgeable about them!

How do I know I didn't learn hallucinations?

Because based on that knowledge, I've produced code that works as intended.

keedaon Qwen3-Coder writes less secure code when told it's for a US persona

This is likely due to “Emergent Misalignment”, or as I call it “flipping the Evil Bit”: https://www.emergent-misalignment.com/

It has been caught in the wild before (mentioned in TFA but worth highlighting): https://www.crowdstrike.com/en-us/blog/crowdstrike-researche...

https://brainbyteslab.org/articles/hidden-vulnerabilities-ai...

Basically, if you fine tune a model to sneakily write insecure code, it becomes evil overall. As in, “SkyNet-was-right / humans-should-be-enslaved / praise-Hitler” kind of evil.

Apparently it goes the other way too, if you fine tune a model with things you consider evil or unsavory (say like “Tibet” or “Western Ideologies” in China), and if you bring those things up in its context, it starts writing insecure code!

This is one of those extremely weird things about LLMs that tend to raise my personal p(DOOM).

keedaon Microsoft and CMU team up to train early career software engineers

Yes, I believe programs like this aim to address this issue too. The traditional ways of learning might no longer be sufficient or reliable enough to produce appropriately skilled workers in the era of AI. Maybe something like this program or other apprenticeship programs would bridge that gap by focusing on juniors who are truly interested in the field.

keedaon Microsoft and CMU team up to train early career software engineers

Yes, that dynamic is essentially what this is proposing to solve. Microsoft is acknowledging that junior engineers will be less in demand due to AI (and presumably it is experiencing that itself) because economics, i.e. when supervised by senior engineers, AI is way cheaper and faster than junior engineers.

These economics would choke off the pipeline that turns junior engineers into senior engineers, which would still presumably needed to supervise the AI.

I don't see the irrationality here?

And yes, bottlenecks at MSFT (or any large corporation) are unrelated to kLOC/hour, which I would argue are primarily because of human factors... which also AI threatens to obviate: https://news.ycombinator.com/item?id=50121996