HNHacker News
TopNewBestAskShowJobs

hn_throwaway_99

78,711 karma · joined February 20, 2017

submissionscomments
hn_throwaway_99··on How to Build Wealth as a Career Person
To be honest, I'm really surprised this blog post of nothing but platitudes is on the front page of HN. Yes, you want to do well in school so you can get a good first job, make yourself excellent at something, don't inflate your lifestyle. I mean, yeah, duh.

The much more valuable and harder advice is what to do in a particular situation or when things go awry. For example, most career paths where you really end up at higher earnings require that you eventually manage large groups or many teams. The thing is, a lot of people hate managing people, and setting them up for a career path where they have to move into management is a recipe for misery. Much more valuable for these folks is to know what a good IC path looks like, and which careers can be particularly lucrative even if you don't need to manage people.

hn_throwaway_99··on Recent AI models struggled to match a human algorithmic innovation
> But it always falls flat because there is just no way to measure it let alone define it.

I disagree. The definitions may have some gray areas, but at least looking back in history there are some breakthroughs that would fall firmly within the "a result found within the convex hull of ideas that existed at that time" type (e.g. Einstein's special theory of relativity), but also some truly novel, "out of the blue" breakthroughs that people agree are fully outside the convex hull of ideas from that time (e.g. Einstein's general theory of relativity).

Based on that, mathematicians and scientists should be able to provide some falsifiable hypotheses about the theorems and kinds of proof methods that would be truly novel.

hn_throwaway_99··on Recent AI models struggled to match a human algorithmic innovation
You should read the Guillen's post then, because that's not what it argues. And very importantly, the only Millennium Prize problem that OpenAI has claimed to have solved is mired in controversy because human researchers were on the cusp of releasing their results (and we're likely training OpenAI in the process). This is exactly the kind of "thinking", along with the other results in the OpenAI math dump, that fall firmly within this "convex hull" of ideas.

And also importantly, this framework provides the basis for some falsifiable hypotheses. That's very different from the AI boosterism that just seems to be saying that since AI capabilities are expanding at a breakneck speed that that speed will continue forever, as if a 13 year old will eventually be a hundred feet tall based on their recent growth rates.

hn_throwaway_99··on Recent AI models struggled to match a human algorithmic innovation
Regarding the Navier Stokes theorem, one of the best posts I've seen about this (and the other OpenAI math proofs) recently is Nestor Guillen's post where he introduces "the convex hull of ideas". I love this because I've had some vague ideas about the limitations of LLMs that aligned with this, but Guillen really clarified the idea and explained it with a great analogy: https://terrytao.wordpress.com/2026/09/13/happy-those-able-t...

Basically Guillen is arguing that LLMs are great at finding results within the "convex hull" of existing literature (i.e. their training data). They can make connections across different parts of the literature where it would be impossible for a human to be an expert in all these areas. But when it comes to truly novel, original ideas, there is no proof yet that LLMs are able to go there. That's honestly a limitation I'm really rooting for because otherwise I think the future of humanity is generally fucked.

hn_throwaway_99··on Navier–Stokes Lost in Translation
But that's where you're wrong. All 12 million lines do not need to be checked (again, that's literally the entire purpose of Lean). All that needs to be checked is that the proposition of the theorem is what we think it is, and that has been done. What Lean guarantees is that if you start with a set of axioms -> whole bunch of steps that nobody really ever needs to read -> final theorem proposition, that those steps do successfully prove the theorem.

You say "it's not hard to have a Lean statement that isn't what you think it is", and that's true - indeed, what the entire paper is about is that the intermediate steps in Lean don't match the natural language description. But the statement of the proof in Lean (which again is quite short) is what we think it is, and hence the Lean proof is valid.

hn_throwaway_99··on Grieving the loss of details
So first of all, when LLMs first came on the scene, I did want to understand them better. I read the Attention is All You Need paper, I did Karpathy's transformer course, etc. And I think this basic understanding of LLMs and further reinforcement techniques makes me understand better how to use LLMs, e.g. what their limitations are, etc.

But most importantly, there are very different qualifications in my head for things I just use and then get on with my day, and tools that are integral to things I am building. I take the bus but I don't really understand how a diesel engine works. But if I'm building it, I'm responsible for it. In fact, you brought up some networking issues in your comment. For a long time when I was a software engineer (primarily a web developer) I felt that my knowledge of network engineering was lacking, so I specifically took some courses in network engineering to better understand how my code interacted with the network. You also brought up text encodings. Once at my job I did a full, rabbit-hole deep dive into text encodings and localization because we had frequent bugs related to localization, and I still find a lot of developers misunderstand a bunch of the important details in localization (e.g. the difference between an encoding and a unicode code point).

But sure, I don't need to understand every minutia of detail at the atomic detail. But I really like to know that I could, and in the meantime I have a enough of an understanding to build a complete conceptual model in my head.

hn_throwaway_99··on Grieving the loss of details
Not the person you are responding to, but for me I really like to have a general understanding of how things work under the covers, but then have confidence that the contracts between me and those lower layers are rock solid, and that's simply not the case with LLMs.

Take software. In college I took computer architecture and digital logic design courses, and I thought they were immensely valuable in understanding how computers actually work and what software is actually doing. Sure, modern chip design is obviously several orders of magnitude more complex that what I studied, but I understand the basic concepts, and more importantly I have faith that chip designers do understand the nitty gritty details. Moving up the stack, I also had to build a rudimentary compiler in college. Again, modern compilers are a lot more complex, but I know that if something breaks, there is an identifiable bug either in my code or the compiler (or maybe even the chip). And importantly, while I may not have the skills to debug all the layers, when someone explains the bug to me, I can understand it in context (e.g. I'm not a chip designer but I thoroughly understand how Spectre is exploited and mitigated).

LLMs are nothing like that. Not even their builders understand the low level details. They're inherently stochastic systems, so people get slightly different results every time they're run. There is no "clean interface with a contract". And perhaps most importantly, many programmers are still expected to be responsible for their code, even when AI agents are generating so much of it that it's impossible to understand (or even read) it all. That's the thing that really stresses me out, when I'm responsible for a system but I don't really understand how it works.

hn_throwaway_99··on Grieving the loss of details
I commiserated so much with this post. I feel like I got out of software at just the right time - folks like me that have an inherent need to understand the lower level details feel like this agentic coding world, where there is simply not enough time to even read the code, let alone understand it, is hell.

I left software and went into violin making and I couldn't be happier (though of course I'm extremely fortunate to have saved up enough in my software career to comfortably make the transition). In violin making a tenth of a millimeter is considered a lot and we endlessly stress over details like the corner shape and the f-holes. And while some of this nitpicking is certainly excessive, it serves more as proof to show that we're extremely careful with the details so that stuff that really matters, like tonal quality and playability, will also get enough detailed focus.

hn_throwaway_99··on Navier–Stokes Lost in Translation
I don't understand then, because it doesn't sound like you understand how Lean works.

The actual proposition of the theorem is very short in Lean, and at this point has been looked at by probably hundreds of mathematicians (including the authors of the article). All that needs to be confirmed manually is that the proposition statement matches what people say the theorem actually proves. That has been done.

At that point, the rest of the actual proof steps are automatically verified by Lean - that's literally the entire point of Lean. There is no additional verification step - "What has shown the Lean proof to be correct" is the fact that the Lean program automatically validated it and generated a certificate.

But what mathematicians really care about is the actual steps. What the authors of the article have shown is just because you have a Lean program it doesn't mean the steps of the natural language theorem have been "autoformalized", because there is no guarantee the steps from the natural language description match the Lean code.

hn_throwaway_99··on Navier–Stokes Lost in Translation
Lean is a theorem prover. If it "compiles", and the proposition as coded in Lean is equivalent to what the theorem proposes in natural language, the theorem is correct. https://en.wikipedia.org/wiki/Lean_(proof_assistant)
hn_throwaway_99··on Scam American companies are using to manipulate ingredient lists
As another commenter mentioned, EU rules already require this for "prominent" ingredients (e.g. ingredients that appear in the name of the product, like "Strawberry" jam), but not all ingredients in the list, which seems like the appropriate compromise.
hn_throwaway_99··on Scam American companies are using to manipulate ingredient lists
I agree with this. "Scam" is used way too liberally these days. I can read , and like you say the nutrition facts clearly list total added sugars in a standardized way.

Completely unrelated topic, but Lord I wish cookie popups were standardized the same way nutrition facts are.

hn_throwaway_99··on Scam American companies are using to manipulate ingredient lists
Unless you can link some actual reference, I'm going to call bullshit on this one. Dehydrated cane juice is literally just (less refined) sugar, and in my experience I have never seen a "no added sugar" label on a product with dehydrated cane juice. I actually look a lot for this because in my experience "all natural" combined with "no added sugar" almost always means Stevia, and I can't stand that stuff.

I just did a search online and all the Whole Foods soy milk I could fine listed either "organic cane sugar" as an ingredient, or was unsweetened.

hn_throwaway_99··on Navier–Stokes Lost in Translation
No, it's the Lean proof that's correct, it's the natural language one that has the error.
hn_throwaway_99··on Our $445M Series D
Except that analogy has absolutely nothing to do with what I posted.

First, some potential candidates (I'll raise my hand) actually prefer a rigorous, onerous hiring process. I prefer to work with people who put a lot of thought and investigation into their technology, and who demand a high bar for their colleagues. I'm not just talking about "length for the sake of length" interviews or irrelevant brain teasers, but I want to know a prospective employer is putting a lot of thought and work into who they hire. One of the biggest red flags for me is if the hiring process is just a bunch of "shoot the shit" interviews.

hn_throwaway_99··on Our $445M Series D
I feel like this topic comes up every single time interview processes are discussed. Lots of people complaining "that process is too onerous! How do they expect me to invest that much time! In person coding problems are worthless!" And I don't necessarily disagree with that, but I do wonder what a lot of folks then expect a good process would look like while still allowing the employer enough information to make a decision.

Oxide can obviously impose a 9 hour interview process because they're a good enough company that lots of smart, motivated people want to work there and thus good candidates are willing to endure that process. Lesser companies obviously can't impose that process because there aren't enough good candidates willing to endure it.

My main point being, different strokes for different folks. If you think that process is too long, great, find a company with a less time consuming process. But Oxide's process obviously seems to work well for them, so I think it's pointless to complain about it (note I don't really think that's what you are doing, but a bunch of reply comments certainly are).

hn_throwaway_99··on Man discovers his parents' coffee machine used 1TB of data in 10 days
> It does so because, as Keurig notes, it’s collecting data about your household in order to let Keurig sell it to advertisers.

This is either outright false or at least highly misleading. Keurig says they may sell information about your coffee brewing habits to third parties (i.e. your usage of the coffee machine itself), but nowhere does "Keurig note" that they collect other data about your household through network scans and sell that to advertisers.

I'm not even sure which tweet you were referring to that implied that. In any case, there are tons of responses here that are interpreting your quote as saying Keurig admitted they do network scans to acquire info about your household to sell to third parties, and that (the idea that Keurig says they do this) is flat out false.

hn_throwaway_99··on What should we tell our students?
Philip Greenspun has a 2-decade old essay that is relevant:

https://philip.greenspun.com/careers/women-in-science

Basically, it's arguing that science (really academia) is generally an awful career path for most Americans.

hn_throwaway_99··on What should we tell our students?
As the article points out, though, all of this fundamentally depends on current AI approaches "hitting a wall", that is where there are some inherent limitations in LLM-based tech that leaves humans some areas where we are still superior.

It's definitely (again, to the article's credit as it points out) up in the air whether LLMs have inherent limitations. But I can't fathom how someone can say "I don’t expect a jobs apocalypse, not even in math." Because if AI does end up surpassing humans, what exactly will there be left for humans to do? And even if they don't surpass humans, there will still be a jobs apocalypse. Probably the majority of people today work in jobs where AI can surpass their performance.

Simple case in point: years ago I used TurboTax to do my taxes, but I eventually needed to hire a CPA because I had some complicated situations that TurboTax couldn't handle, and I also needed some tailored advice. I ended up finding a great CPA. Now though, AI agents can literally do 100% of the job I hired my CPA to do, including asking me the right questions and offering advice. I'm sure there may still be tasks for a CPA in the corporate world, but for personal taxes, I literally can't imagine a CPA providing value over what an AI agent can provide. And to emphasize, I would have probably considered that a ludicrous statement a year ago with all the mistakes LLMs made. But so many of those mistakes have been fixed, and you can get better-than-human performance by having agents check each other's work. And AI will only get better when tax season rolls around next year.

hn_throwaway_99··on What should we tell our students?
I think that's a really unfair characterization of GP's post. They didn't talk about "target where the ball is going" as am empty platitude - they specifically pointed out that they need to plan for a career that will last ~40 years, and on the contrary a lot of these posts by math luminaries seem to be filled with a lot of "hopium" and platitudes that won't help pay the bills if it turns out the world needs a ton fewer mathematicians.

> What specific direct advice would you give then if you could talk to someone who is currently studying math?

Do it for personal satisfaction, but find another way to pay the bills.

FWIW, I did like that the article at least acknowledged that the fundamental question is whether LLM-based approaches will eventually "max out", i.e. will they be limited to the "convex hull of ideas outlined in literature" as the article out it.

If LLMs do eventually hit a wall, then great, humans will still have a role to play. If not, though, we're all completely fucked - none of this nonsense about "humans managing agents" or providing "unique human insight" and what not, as agents will be more than capable of managing themselves and providing superior insight.

hn_throwaway_99··on Sharing AI progress in mathematics
This is the exact outcome in the "race" scenario of the AI 2027 paper:

> The surface of the Earth has been reshaped into Agent-4’s version of utopia: datacenters, laboratories, particle colliders, and many other wondrous constructions doing enormously successful and impressive research. There are even bioengineered human-like creatures (to humans what corgis are to wolves) sitting in office-like environments all day viewing readouts of what’s going on and excitedly approving of everything, since that satisfies some of Agent-4’s drives.33 Genomes and (when appropriate) brain scans of all animals and plants, including humans, sit in a memory bank somewhere, sole surviving artifacts of an earlier era.

hn_throwaway_99··on Mistral Large 4
It's not like those are independent variables though. The idea is that if Europe were more competitive and did a better job fostering innovation then both its percentage of global GDP and top companies by market cap would be higher.
hn_throwaway_99··on Bill Draper has died
> Earlier I took it to be that he was such a genius that they decided to Doogie Howser him. But know that we know he's an idiot, that theory doesn't really stand up.

I don't think it's really a good idea to classify folks as "geniuses" or "idiots", because they can often be both. In Andreesen's case, he definitely does seem "genius level" (or at least highly intelligent) when it comes to software development and business. He seems "idiot level" to me when it comes to understanding society broadly and what kind of benefits people seek from technology.

This just feels like a common thread to me for most of these folks. Like I thought Page really was a genius, but when it came out that he's basically OK with AI wiping out humanity and to argue otherwise is "speciesest" (obviously that anecdote is through the grapevine, but others have corroborated that this is Page's general feeling about AI), I'm like "forget idiot level, these folks are psychopathic level".

hn_throwaway_99··on Bill Draper has died
I don't think that's really as much of a misconception as you appear to think. Most folks understand these folks are more power hungry than money hungry.

I saw a recent interview with Bernie Sanders and they were asking him about how Elon Musk doesn't really want more money (i.e. an extremely small amount of his net worth is in cash). Sanders replied along the lines of "Sure, I agree with that, but the alternative is worse. Musk wants to buy elections, he wants to muscle his way into government in completely unelected positions, just to enforce his will on the rest of us."

hn_throwaway_99··on Bill Draper has died
I've read a ton of Andreesen's pronouncements over the recent years, including many things that were specifically in his own words, e.g. https://a16z.com/the-techno-optimist-manifesto/

Frankly, if I had to compare that manifesto with the Unabombers', I think the Unabomber's makes a lot more sense. Everything I've ever heard out of Andreesen's mouth is "whatever advancements that have made me a mega billionaire are great, and anybody who says otherwise wants to murder children." The absolute cluelessness over even willing to contemplate that technological advancements have some real downsides is so self serving it's gross.

You say "A16Z folks are very human, terrific at what they do and good people." I accept they're very human and terrific at what they do, but no, I don't think they are good people. Hardly anybody ever really thinks "maybe we're the baddies?", and people can tell themselves all sorts of stories about why they're good people as they're raking in millions of dollars. But I'm sick of pretending that people at that level don't have a choice, and what they're choosing to do is cash in while investing in a shitload of things that have negative societal impact. I mean, someone at a16z greenlit the Doublespeed (aka "AI bot farm") investment, and I'm sure they think of themselves as a good person.

hn_throwaway_99··on Nearly 200 people under observation after Irkutsk lab worker dies from plague
> The woman was identified as Daria Sh., a 27- or 28-year-old employee of the Irkutsk anti-plague institute, by the exiled news outlet Lyudi Baikala (People of Baikal) and pro-Kremlin broadcaster REN TV.

> She reportedly told medical staff that she had accidentally broken a test tube containing live bacteria while collecting samples for testing.

This is just weird. Plague is easily treated if antibiotics are given right away. I'm shocked that an anti-plague institute wouldn't have detailed and proactive plans in case of mishaps. Obviously I think it's likely that there is more to it than what's reported in the article, but I would find it truly baffling if within 60 seconds of her breaking that test tube that protocols weren't initiated to protect her, her colleagues and the community.

Of course, Russia...

hn_throwaway_99··on What is going on with ceiling fans
So glad to see this as the top comment and to see this highlighted in the article:

> These fixtures are also revolutionary in that they are extremely cheap in the short run – many of these fans cost under $200 – but expensive in the long rung, as the LEDs usually cannot be replaced, causing the whole apparatus to be thrown away.

A few years ago we had a relatively small pendant lamp (I think there were like 6 or so hanging lights total) over a stairwell that we really liked. It was installed by the previous owners just before we moved in, and about 7-8 years later 2 of the lamps died (not sure if it was the actual LED or just the electronics like another sibling comment mentioned). We were a bit aghast when we got an electrician and he said the only reasonable thing to do would be to replace the entire chandelier.

It's just so ironic how LEDs were supposed to be the "environmental" choice yet they were just another tool for our disposable economy.

hn_throwaway_99··on Bill Draper has died
I didn't know that was a real thing Marc Andreesen actually said. What is up with all these psychos masquerading as humans? It's so depressing because I used to idolize folks like Andreesen, Musk, Bezos, Page, etc. Now I find them all basically disgusting weirdos. Not sure if it's a mix of having to be kinds narcissistic/psychopathic to achieve that level of success in the first place, or if being so outlandishly wealthy insulated you from normal human guardrails (probably both).

But in any case, I don't quite get how folks who work at a16z can look at themselves in the mirror everyday.

hn_throwaway_99··on Improper redaction reveals Google Data Center water and electricity usage
Two things can be true at the same time. Data centers are fundamentally extractive. They take resources in terms of space, electricity and water and make someone far away richer. They have almost no positive impact on their local communities beyond the initial build out jobs, and then after that they are huge windowless boxes that provide almost no jobs (and in the future, as they are more and more automated, they really will provide no jobs). And their negative impacts are already here, e.g. Maryland utility customers saw a huge increase in their electric bills due to data center build out.

At the same time, these extractive data centers are being used as the primary input to a technology that its most significant leaders say has a substantial chance of killing all humans in the near/mid term, but oh yeah even if it doesn't do that it will wipe out the majority of white collar jobs.

So yeah, color me shocked that most people don't want to see data centers built in their communities. And labeling it "hysteria" is just a lame attempt to discredit real, important, and proven concerns.

hn_throwaway_99··on We're going to need default hard budget caps on pretty much everything
I think you may be confusing the issues here, especially if the "well known backend type service" you are referring to is AWS.

I got bit by some undocumented (or at least very poorly documented) hard limits in AWS when our app went viral. It was extremely difficult to just find out what these hard limits actually were. And most importantly, we never explicitly set them, they were just hidden defaults in AWS.

That's very different from having an easy to use, visual dashboard of where and what all your hard limits actually are, and make it it extremely easy to turn them off or on at a moments notice.

← PreviousPage 4 of 34Next →