HNHacker News
TopNewBestAskShowJobs

blagie

5,691 karma · joined March 22, 2018

submissionscomments
blagie··on Intel Arc Pro B50 GPU Launched at $349 for Compact Workstations
Well, no. It doesn't. The comparison is to the A1000.

Toss in a 5060 Ti into the compare table, and we're in an entirely different playing field.

There are reasons to buy the workstation NVidia cards over the consumer ones, but those mostly go away when looking at something like the new Intel. Unless one is in an exceptionally power-constrained environment, yet has room for a full-sized card (not SFF or laptop), I can't see a time the B50 would even be in the running against a 5060 Ti, 4060 Ti, or even 3060 Ti.

blagie··on Poisoning Well
The whole design was based on things to make pages not discoverable except to sketchy AI companies who violate web norms.
blagie··on Poisoning Well
"want to exploit"????
blagie··on Poisoning Well
It's competition. Poison increases in toxicity over time.

I could generate subtly wrong information on the internet LLMs would continue to swallow up.

blagie··on Poisoning Well
I'd be totally down with "content belongs to everyone."

The problem is when you steal my content, repackage it, and resell it. At that point, my content doesn't belong to everyone, or even to me, but to you.

* I'd have no problem with OpenAI, the non-profit developing open source AI models and governance models, scraping everyone's web pages and using it for the public good.

* I have every problem with OpenAI, the sketchy for-profit, stealing content from my web page so their LLMs can regenerate my content for proprietary products, cutting me out-of-the-loop.

blagie··on Sam Altman Says Meta Offered OpenAI Staffers $100M Bonuses
The curve is getting steeper, yes. That's not a contrast to the "AI will replace programmers" rhetoric.

Steeper means: higher at the top. Lower on the bottom.

Right now, AI can do the job of the bottom large percentage of programmers better than those programmers. Look up how a disruptive S-curve works. At the end, we may be left with one programmer overseeing an AI "improving" itself. Or perhaps zero. Or perhaps one per project. We don't know yet.

Good analogue is automation. Mass-scale manufacturing jobs were replaced by a handful of higher-paid, higher-skilled jobs. Certain career classes disappeared entirely.

blagie··on Tracking Copilot vs. Codex vs. Cursor vs. Devin PR Performance
> If entire commit is generated by AI then it is obvious what created it - it’s AI.

This is not the case. The output of a compiler is 100% created by a compiler too. Copyright is based on where the creative aspect comes from.

I have had very little luck having 2025-era AIs manage the creative aspects of coding -- design, architecture, and similar -- and that's doubly true for what appears to be the relatively simplistic model in codex (as far as I can tell, codex trades off model complexity for model time; the model does a massive amount of work for a relatively small change).

However, it is much better than I am at the mechanical aspects. LLMs can fix mechanical bugs almost instantly (the sort of thing with a cut-and-paste fix in some build process from Stack Overflow), and generate massive amounts of code without typos or shallow bugs.

A good analogy is working with powertools versus handtools. I can do much more in one step, but I'm still in creative control.

The codebase I'm working on is pretty sophisticated, and I might imagine they could implement more cookiecutter things (e.g. a standard oauth workflow) more automatically.

However, even there -- or in discussions with larger models about my existing codebase -- what they do is in part based their creativity on human contributions to their training set. I'm not sure how to weigh that. An LLM oauth workflow might be considered the creative median of a lot of human-written code.

I write a lot of AGPL code, and at least in the 3.5 era, they were clearly trained on my code, and would happily print it out more-or-less verbatim. Indeed, it was to the point where I complained to OpenAI about it at the time, but never got a response. I suspect a lot of generated code will include some fractional contribution from me now (an infinitesimal fraction most of the time, but more substantial for niche code similar to my codebase).

So in generated code, we have a mixture of at least a few different pieces:

- User's contributions, in prompt, review, etc.

- Machine contributions

- Training set contributions

blagie··on OpenAI slams court order to save all ChatGPT logs, including deleted chats
Because such cases don't really exist.

Your text used an unusual double ellipsis (" .. " instead of "... "), uncommon (not rare) generative vocabulary ("panacea"), etc. Statistics on those allows for pretty good re-identification.

Ditto for times you do things and work schedule.

Etc.

It's not "obviously identifiable," but a buffer overflow is not "obviously exploitable." Rather, it takes a very, very expert individual to write a script before everyone can exploit it.

Ditto here.

blagie··on OpenAI slams court order to save all ChatGPT logs, including deleted chats
The entire mess isn't with data in databases, but on laptops for offline analysis, in log files, backups, etc.

It's easy enough to have a SQL query to delete a users' data from the production database for real.

It's all the other places the data goes that's a mess, and a robust system of deletion via encryption could work fine in most of those places, at least in the abstract with the proper tooling.

blagie··on OpenAI slams court order to save all ChatGPT logs, including deleted chats
This data cannot be anonymized. This is trivial provable, both mathematically, but given the type of data, it should also be intuitively obvious to even the most casual observer.

If you're talking to ChatGPT about being hunted by a Mexican cartel, and having escaped to your Uncle's vacation home in Maine -- which is the sort of thing a tiny (but non-zero) minority of people ask LLMs about -- that's 100% identifying.

And if the Mexican cartel finds out, e.g. because NY Times had a digital compromise at their law firm, that means someone is dead.

Legally, I think NY Times is 100% right in this lawsuit holistically, but this is a move which may -- quite literally -- kill people.

blagie··on Precious Plastic is in trouble
I will mention. Ocean Cleanup has a 55 million Euro budget. Precious Plastic is being criticized for having squandered a 100k donation. That 550 times less.

Ocean Cleanup claims to have removed 21,000 metric tons of plastic in its existence.

If we assume linear growth in budget, that's $330M Euro / 21k ton = 15,000 Euro per ton of garbage.

About 10 million tons of plastic end up in the ocean each year, so over their history, they've cleaned up 0.2% of the annual input.

I can't really comment if any of that good or bad, but I thought numbers (even if squishy estimates) would be helpful to inform the discussion.

blagie··on Precious Plastic is in trouble
I agree with almost everything, except that it's a problem if someone donates.

Some of the donation game is a competition (for example, applying for grants, or foundation funding), where the dynamic you described dominates. Much of it is not. In many cases, a rich person just cuts a check out of their leisure budget. That's why I compared this to buying a supercar. The rich person learns something, and has something fun to do more constructive than, well, many of the things rich people do.

Money is a social construct; it's just a way of keeping score, and organizing people. At the end, what matters is what you do, and what resources you use. The better question isn't what else the /money/ could be used for, but what else the actual inputs --- the /people, space, and tools/ --- could be used for.

If those same people are making military weapons, optimizing ad clicks, or running cons, that's a negative use of resources.

It's very possible a rich person decides they just want to:

- give a bunch of people space to follow their passion (same as an arts grant);

- view this as a part of personal development (same as giving tuition to a college);

- a research grant (interesting open information will come out, which is perhaps a few steps away from being useful;

- promotion of recycling; or

- just funding this on the off-chance something big comes of it.

It's all good.

blagie··on Precious Plastic is in trouble
"Con" is a strong word, and I'm not sure appropriate.

If someone was taking a $500k salary from donated funds, I'd feel it's a con. I see nothing to that effect.

People without a clue sometimes pick up a clue after some time. Donors donate for a variety of reasons. There are angel investors who donate simply because they want to push a concept along, or want to help nice people.

If someone wealthy decides to support them instead of e.g. buying a supercar or a $10M painting, I think it's all good.

Live and let live.

blagie··on Mathpad: A mathematical keypad for students and professionals
I'd pay money for one of these, if it had the keys which I want, which this one doesn't. I think the trick would be to somehow make it customizable, but I'm not sure the best way to do that. Parametrized PCB design? 3d printing?

I'm not quite sure, but my dream would be smaller buttons, coarsely modelled on the HP48-style keypad. I can't imagine many people share that dream, but I've wanted to make that for a long, long time.

Coincidentally, you get a long ways there by simple adding a Greek keyboard layout. You hit meta-space, and ασδζχψ. meta-space, and back to normal.

Foreign-language keyboard under GNU/Linux also have a layer with many convenient math keys (²·≈·¾ all exist natively in my normal layout).

A smaller win for me would be to simply print all of that on my keycaps.

blagie··on Google AI Ultra
I'm holding out for "Ultra Max Pro."

(Comment is on the horrible naming; good naming schemes plan ahead for next month's offerings)

blagie··on InventWood is about to mass-produce wood that's stronger than steel
Drop a Google Streetview in the middle on downtown classical Gothenberg, and try to find a wood construction building. Sweden is about as northern Europe as it gets. Head over to Amsterdam, and it will look pretty similar.

It will be relatively similar to most other major European cities; solid masonry construction has historically been the norm.

You'll find some wooden houses in smaller towns and rural areas, but even there, if you head over to e.g. Poland, most things will be solid masonry. And the wooden construction is mostly relatively new; what you'll rarely find is old wooden construction.

blagie··on InventWood is about to mass-produce wood that's stronger than steel
Why? That seems dumb.

We have concrete buildings going back to Greek and Roman times, mostly standing.

Wooden buildings have a very finite lifespan, and, aside from earthquakes, much less so in a disaster. A wooden city and a steel/brick/concrete city react very differently to firebombing.

Especially with declining populations, it makes sense to build structures to last. The old school brick/concrete/stone/... European buildings from 1500 are largely in good shape, at least where they weren't demolished by war.

Traditional European construction is also much more friendly for heating / cooling. The thermal capacity of stone-based materials helps a lot.

To me, it feels a bit like Germany banning nuclear power plants, and switching to Russian fossil fuels instead...

blagie··on The Future Is Too Expensive – A New Theory on Collapsing Birth Rates
It's going to become acute well before. It's more like the tragedy of the commons.

You need a tax / working base of some size to produce enough stuff. That simply won't be in place when the current kids are adults, and adults or elderly.

Barring robots and AI -- which admittedly looks like a very real possibility now -- the current adults will not have enough working-age people to support them when they're old.

Indeed, the problem is acute only because they won't be dead. We're all counting on our kids to support us when we're old, only now indirectly, via taxes and mechanisms like social security.

If South Korea simply had 1/4 of its current population, everyone would have more space and resources, and things would probably be okay. Property values and economy would contract in absolute terms, but everyone has more stuff.

blagie··on The Future Is Too Expensive – A New Theory on Collapsing Birth Rates
No, the population has not "levelled out" in Korea.

These dynamics have a pretty long timeline. To give an extreme example, if you, for example, have a city with 10M old people who can't have kids anymore, you can reliably predict the population will be 0 once they pass.

The demographic pyramid of South Korea is scary:

https://en.wikipedia.org/wiki/Demographics_of_South_Korea#/m...

Learn how to read these plots and what they mean.

There are almost no young people.

There are almost no kids in South Korea.

Unless there is mass immigration, the population of South Korea will implode.

Even if people began breeding like rabbits today, there is nothing that can be done; when the 0-20 year old population is of working age, there will be a massive number of old people, and almost no one of working age.

blagie··on VPN firm says it didn't know customers had lifetime subscriptions, cancels them
I'll give another lens for why lifetime subscriptions often make sense. Places to consider them:

* Non-profits

* Clubs

* Academic organizations

* Educational / semi-educational companies

* ...

The key thing to remember is that the whole of the universe isn't transactional.

And the right thing to do for "an [organization] in need of cash," if you'd like to see them continue, might be to... give them cash. It might also be the right thing if you'd like an organization to be able to bootstrap and not be under the pressure of investors. There are fine reasons to keep some organizations private and customer-funded (especially if the founder has a strong moral backbone).

blagie··on Burrito Now, Pay Later
> You argue that housing has a finite supply, but that's true of cars and steel and everything: there is always a finite supply

Land is a finite supply. With zoning (and with exponentially rising costs for skyscrapers over one-story homes), so is housing on land. At least in places where housing is expensive (e.g. major cities like SFO, NYC, Tokyo, Shanghai, etc.), this is a limited, rent-extracting resource.

The Earth's crust is about 5% iron. Cars are limited by our capacity to produce them.

> I know that you mean competition from not-private providers, but that's not explaining why a not-free provider is good.

No, that's not what I mean. I would recommend taking a basic econ course to understand jargon like "free good," "scarce good," and "rent." I won't fix that in a web post.

Khan Academy is not bad for this kind of basic economics.

> So can you, but you are the one making exceptional claims.

Ah yes. The great controversy of Adam Smith and basic economics.

I gave one link, which you clearly didn't bother to even click through.

> Links?? And no, I'm not going to look it up. You look it up.

Fine. Wallow in ignorance. I won't be responding further.

https://xkcd.com/386/

blagie··on Burrito Now, Pay Later
You misunderstand both my premise and basic economics.

I would recommend reading the original Wealth of Nations by Adam Smith, which forms the basis of modern capitalism.

https://www.gutenberg.org/ebooks/3300

It's clear, analytical, and mostly correct.

Chapter XI, The Rent of Land, is pretty quaint at this point, but gives a compelling counterpoint to your argument. It begins:

"Rent, considered as the price paid for the use of land, is naturally the highest which the tenant can afford to pay in the actual circumstances of the land. In adjusting the terms of the lease, the landlord endeavours to leave him no greater share of the produce than what is sufficient to keep up the stock from which he furnishes the seed, pays the labour, and purchases and maintains the cattle and other instruments of husbandry, together with the ordinary profits of farming stock in the neighbourhood. This is evidently the smallest share with which the tenant can content himself, without being a loser, and the landlord seldom means to leave him any more."

The basic analysis of the chapter can be translated from 1780-era industries into 2025-era industries, and explains much of how housing markets work.

If you would like contemporary data, I would recommend looking at the impact on the price of housing of European cities which offered nationalized below-market housing.

A central fallacy in most of the libertarian space stems from the difference between industries which are not zero-sum (e.g. manufacturing, software, writing, etc.), and rent-collecting ones (e.g. literal rent, much of finance, mining, etc.). Here, I use "rent" as economics jargon, not common meaning. You can look it up.

Being used to operating in domains like software -- which has literally zero marginal cost -- there is a natural tendency to want to extrapolate the same dynamics to other industries.

However, economics works out differently in industries like that. For the most part, if there is a finite amount of stuff, it makes sense to divide it fairly to avoid dynamics like that. If you're producing stuff, it makes sense to divide it according to how much you contribute.

blagie··on Burrito Now, Pay Later
I'm not so sure. I can describe some of the paradoxes in our social systems in-depth, but it's often the poorest people saying we can't afford social improvements.
blagie··on Burrito Now, Pay Later
A major goal of policy is to set incentives to which people optimize.

The "capitalist soul sucking robber barons" are doing exactly what the incentives we set up push for. If they didn't, boards of directors / shareholders would vote them out. And if those didn't, competing businesses would crush them. No one has any real choice; it's the free hand of the market.

There are nice books about this (Dictator's Handbook comes to mind).

Many would prefer to be less exploitative personally, but need legal frameworks to be able to do so. And indeed, in history, you often saw the "capitalist soul sucking robber barons" see the harms they were causing and push for that regulation. Most people would prefer to provide safe food, safe medicine, etc. than sell poison and snakeoil. But you can't really resist market forces.

So you need to work to set up socially-constructive market forces.

And it's all better than serfdom and 1500-era systems regardless.

blagie··on Burrito Now, Pay Later
It's a little bit different. "The projects" were only available to the very poor, and often by lottery. They often turned into a ghetto.

I'd like a baseline option open to everyone.

If you just finished Stanford, and want to bootstrap a startup, teach kids, or do a non-profit, you should be able to move into a "project" too while getting things off the ground. If you're a student and want to save a bit of money, you can move in too.

If you want to upgrade at some point, great.

If not, you've got a safe bed to sleep, and a roof overhead.

Part of the difference, too, is if open to everyone, voters can push for improvements. A lot of what went wrong with the projects wasn't due to a lack of resources, but due to mismanagement of resources. For nationalized goods to be well-managed, voters need a stake. A more extreme example of this is prisons.

But the critical thing is at that point, for-profit housing has competition. I might gladly spend $500/month to have better housing, but I might not be willing to spend $5000/month (insert your own numbers for single/family/income/etc.). That prevents some of the most exploitative practices.

blagie··on Burrito Now, Pay Later
It depends on the format of mortgages.

There are parts of the US where home ownership is bottlenecked by land prices, where this is absolutely the case.

And there are parts where it's bottlenecked by construction costs, where this is not the case.

"Are a bad thing" is also complex. I would agree they WERE a bad thing when introduced.

Today, there is a complex problem of how you unwind this phenomenon. If you invested your life savings plus $800k of debt into a $1M home, and it's worth $500k, you have a serious problem. Simply eliminating access to debt would cause all sorts of economic disruption for consumers, almost certainly outweighing the benefits in lower housing costs.

And the other piece of the puzzle are zoning and codes. 100 years ago, if you had a plot of land, you could cut the wood and build a log cabin. Today, construction is $$$$$$$ because of inspections, licenses, insurance, and a crew of bonded / licensed workers. A log cabin would be illegal and result in fines.

blagie··on Burrito Now, Pay Later
Personally, I think we should have all rent-extracting industries nationalized.

Land, like in China, should be available only under a 99-year lease (perhaps with a modest homestead exemption). The government should own radio airwaves, natural resources, and banking. This doesn't mean that it couldn't hire private contractors to do the work (e.g. bidding on oil extraction), but Chevron and Exxon shouldn't own the oil field.

And I think the government should provide very basic, minimal, Soviet-grade housing free to everyone.

If you'd like your kids to have their own bedrooms, you pay. If you're happy with a 2-bedroom for a family of four (kids room, adults room), or dorm-style housing for adults, you can live there for free.

Having a free option would, in fact, drive costs down for everyone.

More broadly, we should rationalize what's nationalized. For example, a lot of parts of education make sense with a lot more open competition.

blagie··on Burrito Now, Pay Later
I don't know that's quite true.

I've seen cases where it allows those with money to extract money from those with a lot of money too.

Take this situation:

Business A and business B are both doing well and competing with each other in a duopoly.

a) No credit: Each business needs to be in the black, and competition is sustainable.

b) Credit: Whichever business borrows more money to spend on buying customers is more competitive in the short term (which can lead to monopoly long-term). Both businesses run in the red are and brittle. In any downturn, tightening of credit, etc. both are liable to go under.

This competition to raise increasing amount of money to be competitive (to increasing devaluation and/or debt makes for a pretty broken economy), and makes traditional businesses (which invest in R&D from profits) increasingly unsustainable.

This even goes to government level. Historically, if e.g. two European powers were in a war, whomever could borrow more to buy weapons / hire mercenaries would have an advantage. Access to credit made for more costly, more deadly wars, and broken national economies. If neither side had access to credit, both sides would be strictly better off.

blagie··on Evolving OpenAI's Structure
That's the reality of every organization if it survives long enough.

Checks-and-balances need to be robust enough to survive bad people. Otherwise, they're not checks-and-balances.

One of the tricks is a broad range of diverse stakeholders with enforcement power. For example, if OpenAI does anything non-open, you'd like organizations FSF, CC, and similar to be represented on their board and to be able to enforce those rules in court.

blagie··on Evolving OpenAI's Structure
To be specific: The nonprofit currently retains control. It will stop once more dilution sets in.
← PreviousPage 4 of 34Next →