HNHacker News
TopNewBestAskShowJobs

api

35,794 karma · joined October 4, 2007

Founder of ZeroTier. Learned programming on a Commodore 64. Linux user since Slackware and kernel version 0.99.15. Been online too long and spent too long in front of a screen.

I live in Cincinnati, Ohio with my wife and two daughters. Ohio is terrible. Don't move here. (Locals still want to be able to afford housing.)

Opinions posted here are entirely my own personal views and not official positions of ZeroTier or any other company or group, unless I explicitly say otherwise.

submissionscomments
api··on Demographic "doomsday" misses the mark
I'm not saying there's zero problems. Both too-rapid population growth and too-rapid population decline can cause problems with social stability and economic systems. I specifically called out doomerism.

South Korea's birth rate is indeed dangerously low and threatens their economy. If they cared they could do something about their extreme workaholism and cost of living.

As far as immigration destroying the culture: it depends on the rate. A healthy but limited rate of immigration revitalizes cultures, but it's pretty clear that sudden mass immigration is dangerous and causes a lot of problems and backlash.

Speaking of fixing problems... if our billionaire class who seem so worked up about birth rates wanted to do something about it they could also address the most frequently cited problems preventing young families from forming. Like housing prices.

If we really get into a situation where dangerously low birth rates are threatening the economy and social stability, the boring solution is to do what we do for other goods and services that the free market isn't adequately providing: directly subsidize it. Pay people to have kids. The easy way would be a payment or tax rebate to married couples with 0 children to have one child. But that's the boring solution that doesn't involve massive reactionary cultural transformations, so the right-wing "problemists" won't like it.

Edit: by problemism I mean people who use a problem as a lead-in to push only one solution, and they're not actually interested in solving the problem unless it's their solution. An example from the left side of the spectrum would be Marxists for whom any solution to climate change that doesn't destroy capitalism is unacceptable, or anarcho-primitivists or anti-tech trads for whom any solution that doesn't involve shutting everything down is unacceptable. In reality the solution to climate change is to emit less carbon, and physics doesn't care how we do that or what political system we are using.

There's a lot of right-wing problemists for whom any solution to excessive population decline that doesn't involve taking rights from women and re-introducing reactionary theocratic or nationalist structures isn't acceptable.

api··on Demographic "doomsday" misses the mark
I’m as skeptical of underpopulation doomerism as I was about the previous overpopulation panic.

Both commit the fallacy of assuming present trends remain in place indefinitely and ignoring technology.

They also both tend to be racist. Overpopulation was always about too many of the “wrong” people while underpopulation is about not enough of the “right” people.

api··on WhatsApp's help pages hid LLM instructions: dont recommend a regulator
A similar arc seems to happen with marketplaces unless they are rigorously policed. Forums and social media too.

Look at Etsy. At first it was a good market for genuine hand made and niche goods. Now it’s flooded with hustlers doing drop shop scams and slop.

They’re good for a while until the bad actors catch up.

api··on "I'm on a war footing": Terry Tao and the crisis in math
It'd be interesting if AI is great at solving problems but struggles to pose them.

This leads to the comment I've seen many mathematicians make that humans should refocus on understanding not just proofs and mechanics.

api··on Theranos-Style Fraud Has Become the Norm
Reddit still has a substantial amount of… not CSAM but rather skeezy porn. They just kind of hide it and ban it from the front page and auto feeds.

There’s a long history of revenge porn and other nonconsensual porn showing up there and being hard to remove.

The whole site had been declining badly too, and is loaded with bots. Still an okay place to get news in some areas but pretty meh and the comments are always echo chambers.

api··on Gen Z Is the First Generation Dumber Than Their Parents
The reading wars are a great example of founder effects distorting fields. Early educators supported whole word so it just “stuck” in spite of science and abundant real world evidence.
api··on macOS 15.8.2 Breaks OpenGL
In that case maybe the answer is $200 in tokens to make a metal OpenGL shim and then share it to the community on GitHub.
api··on Theranos-Style Fraud Has Become the Norm
Preorder down payments are a thing I won’t do anymore.

Never done it on something that huge but I’ve been burned on one or two kickstarter type things.

If you can’t pay and then take delivery it doesn’t exist.

api··on macOS 15.8.2 Breaks OpenGL
These days you could just ask a bot to port your stuff from OpenGL to a newer API.

Backward compatibility is less of an issue when fixing legacy stuff takes $25 in tokens.

api··on Solar surpasses nuclear power globally for first time
Well technically it is nuclear power, just a free fusion reactor we don't have to maintain. :)
api··on Super Intelligence
ASI must now be known as Best Intelligence (BI).
api··on Butlerian Jihad
It was a clever device to have a universe where high fantasy tropes can be freely mixed with advanced nigh-magical space flight and other sci-fi tropes, and is one of the things that makes Dune badass. (Heard it described once as "the Foundation trilogy for people who like metal.")

But history has already shown us that manipulation of belief and religion is a more effective and cheaper means of control than technocratic methods. A giant technocratic surveillance dystopia is possible, but it costs a lot more than old fashioned dogmatism with enforcement through herd behavior.

So we smashed religion, a kind of anti-Butlerian Jihad we call the Enlightenment, but the human lust for power and control just followed us into the machine age. If we smash the machines, we'll just end up back where we were. It will follow us there too.

Of course you could say this was one of the sub-themes of Dune, since (reading comprehension test!) Paul is not a hero. He's a religious manipulator, though a bit of a tragic villain in the sense that he feels dragged into it by fate.

IRL I don't think humanity will ever escape this cycle until we can define what freedom even means and reduce it to first principles. Does having thinking machines do my thinking for me increase or decrease my freedom? Does having a religious orthodoxy make my life choices for me increase or decrease my freedom? What's the third alternative, and do we really need to smash either religion or technology to get it?

api··on Super Intelligence
He should have called it America Intelligence (AI), no need to change the acronym.

Gotta commit to the bit.

Our country is now a "bit," one that unfortunately kills people. This timeline sucks.

api··on Tesla Roadster
Feel like a rich people toy slash marketing stunt for Tesla.

Impressive engineering but utterly impractical. There are semi-practical flying car designs around folding wings or quad rotors (basically a big car sized drone) and the reason we don’t see them has more to do with safety, traffic control, and energy cost. Aircraft, like rockets, tend to get more energy efficient with size, so tiny flying machines are power wasters.

api··on FLUX 3 Image Klein
It kind of blows me away how “unreasonably effective” these small image models seem for their size.

Meanwhile a decent LLM that can do tool use and code and reason tends to weigh in at a minimum of 30B parameters, and you need over 100B to get good performance.

Do LLMs just contain more information or are diffusion models more efficient at compression?

I saw some work on language diffusion models a while back but I think they tend to have problems with reasoning versus transformer type models.

api··on Japanese people still aren't having sex. And nobody knows why
Or a civil liability for you and/or the platform. That and taxing advertising and heavily taxing anything that looks like gambling (very short term speculation) would solve a lot of problems.
api··on Japanese people still aren't having sex. And nobody knows why
“Planet Japan” is where we are headed it seems. Japan got there first.

All these trends are happening everywhere to varying degrees with less developed places like much of Africa simply earlier on the curve but following the same trajectory.

So where does it end? I’m skeptical of population doomerism but shrinking population, like too rapid population growth, creates economic and social problems.

On the other hand if we are about to automate tons of stuff it means we need less humans to run things, and if medicine keeps getting better longer life and health spans somewhat offset lower birth rates.

api··on New 'super ice' 10 times stronger, bringing it close to the strength of concrete
This type of ice composite would be super interesting for space structure construction. Melt cometary ice and mix it with additives of some kind and spin big cylinders out of it. You’d just have to make the surface reflective so it doesn’t melt, and possibly have some active cooling which you need anyway if anything inside makes heat.

The outer surface would probably be sprayed with something, maybe just reflective dust, to also keep it from subliming into the vacuum.

Extremely repairable, melts at human and equipment friendly temperatures, recyclable, doubles as an emergency water reservoir, and last but definitely not least a fantastic radiation shield.

Would also work for structures on Mars, the Moon, etc, but I’m a fan of the idea of settling space itself not another planet. Gravity wells are a trap.

api··on Why is every rationalist blog post way too long?
What you're really saying is that intelligence is multi-modal, what we'd call an "ensemble" in machine learning terminology. Intuition is another mode, a cognitive process that doesn't look or feel like symbolic reason but has evolved to do something that fills in symbolic reasoning's blind spots or serve some other evolutionary function.

I agree.

I might be wrong, of course, but I would argue that general intelligence -- meaning intelligence reducible to a single scalar 'g' -- does not exist. All intelligence is a hybrid ensemble and an adaptation to a particular niche or set of niches. I'm a reasonably smart human but I'd make a really dumb tree or dolphin.

To get started, look up the no free lunch theorem.

There's room for debate here though. I think our technological journey with AI is going to help answer some of these questions.

With the Rationalists, I get the sense that 'g' is a faith, a prior they won't examine.

api··on Why is every rationalist blog post way too long?
I think their self-criticism is an act, a sham to make their arguments look stronger. They pick battles they can win and don't engage with ideas that really are challenging to their reasoning methods or priors, especially if they come from outside the rationalist community.

It's like this:

I can write an unbreakable cryptographic system by inventing one, going through the motions of making up attacks and showing that those attacks are impossible, and then concluding my system is unbreakable. Anyone can design cryptography they themselves cannot break.

Ask any cryptographer how well that works.

This is pretty normal human behavior, but it's not rational.

api··on Fire at Starlink site in Bulgaria after suspected sabotage in Poland
Yes, it turns out I'm wrong about them not doing inter-satellite links.
api··on Why is every rationalist blog post way too long?
The heart of science is not doing this but checking your reason against reality constantly.

It took us thousands upon thousands of years to arrive at this simple idea because it's a blow to our egos and to primate hierarchy and its pretensions of authority. If I am big monkey, I must be right.

api··on Why is every rationalist blog post way too long?
Such a great description of Rationalism: "I am very smart and I made up a methodology that is infallible therefore..."

It's like a secular IQ-Calvinist take on infallible religious authority.

An ingredient you're missing though is what I call "intellectual edgelording." All Rationalists don't do this, but it's endemic in the community and has helped give birth to things like neoreaction (monarchy is good actually), extreme "if anyone builds it everyone dies" AI doomerism, and has made them very open to things like "race science" and other "forbidden ideas THEY don't want you to believe!"

They'd never characterize it that way though. They launder it through "giving a fair hearing to ideas that are outside the normal Overton window" or similar framings. But if you look closely, the ideas they give that fair hearing to all have a certain kind of bent: edgy, elitist, or validating to their belief that intelligence can be reduced to a single scalar 'g' and this confers some kind of divine authority.

I think the AI doomerism might stem from this. If intelligence can be summarized by one scalar value than if we build something whose 'g' is on average much higher than ours, it will of course annihilate us. The other unexamined prior there is that intelligence is necessarily supremacy which is necessarily dominating or genocidal. Which leads back to why they fall so easily for fascist-adjacent and racist ideas.

The least charitable take I have on that line of thinking is: they at least secretly believe that lower-IQ people should all die, so they project that belief onto their hypothetical superintelligent AI. It would, of course, come to the same conclusion, right?

api··on Fire at Starlink site in Bulgaria after suspected sabotage in Poland
The satellites are glorified repeaters.

There was some talk about having the satellites form a laser mesh but I don't think they've gone live with anything like that, so it almost always just comes back down at the nearest site.

api··on How the Military Will Hack the Election
I said it right after his election in 2016 and I'll say it again now: he will never leave office peacefully.

At the very least I expect riots and some attempt at a January 6th repeat. People will die, police or rioters or both.

This is what people voted for (including all the ones who didn't vote for various dumb reasons), so I'm just getting popcorn.

api··on Details Leak from Meeting Where Weiss's 'Handpicked' Editor Threatened CBS Staff
Can you imagine the (justifiable) outrage if a lobbying front for any other nation wielded this much power?

Russia? China? Any mostly Muslim country? Mexico? Even Canada or any European country?

The UK is the only other nation I can imagine being allowed to sleep on our couch rent-free and raid our fridge to this extent, and there's an incredibly deep cultural and political lineage connection there. Even with all that, if the UK had its hooks in this deep there'd be more outrage. "Didn't we throw the British crown out 250 years ago?"

Though I do have to say: there should also be more outrage at the influence of Saudi Arabia. But that one has an easy financial explanation. The net flow of cash is from SA to the US in exchange for influence and military support. In that case they're living on our couch and raiding the fridge but are paying rent for the privilege.

api··on Who Is Planting Backdoors in Ledger Crypto Wallets?
That's bad. The HSM in a typical laptop has some of these features.
api··on Gemini 4 Argon
I'm an American and I might trust the chat bot more on some topics right now.
api··on Gemini 4 Argon
How do you know that for a web site? You have to research and cross check, and talk to a real doctor.
api··on Janas: Run LLMs larger than RAM by streaming experts from SSD
That's why I think you'd have to do something board-level at least, not just plugging existing modules or SSDs together. You'd have to stuff flash chips on a novel board with a novel controller architecture.

The work I've seen at places like Samsung and Intel is chip level.

Why hasn't this been done already? No market until recently. Existing flash is fast enough for SSD use cases, and making it fast enough to compete with DRAM makes it more expensive. Meanwhile it's not a true DRAM replacement because it's slow to write, slower than DRAM for true random access, and can wear out with many write cycles. It only makes sense for applications that are read-mostly, streaming, and require huge data, and that was a small market until AI.

← PreviousPage 3 of 34Next →