GuB-42

Your comment about cartels, monopolies, etc... actually support my point that when people say that dynamic pricing is a problem, usually, the problem is elsewhere. Monopolies are a problem and something economists recognized (though it is not as simple as "monopolies bad"), and there are laws against it.

Free market is indeed somewhat fictional, as it is far from free in practice. The stock market in particular may look unhinged, but it is heavily regulated, that's the reason why it works, and the reason it looks like the ideal of a free market is, ironically, because of regulations.

As for diamonds, these are luxury goods, being expensive is the whole point, so what if they are overpriced? If de Beers didn't have a history of things like using slave labor, I wouldn't have a problem with it. Would I have hated it, if my $10k diamond turned out $500, yes, no one likes losing money from a bad investment, but prices have to go down at some point. It also shows that competition can work, de Beers didn't go down because of regulation, it went down because it was beaten by technology and bad publicity.

The problem with Shkreli and the Epipen was because healthcare is not a normal market and completely unlike diamonds. Normally, one would expect the government to make sure that life-saving drugs stay affordable, and it is the case in many countries, the US being a notable exception. Instead, in the US, the government promotes this kind of behavior though a combination of its mostly private health insurance system, tight control on who gets to make and sell drugs, but no control on the price. EU countries (most of the developed world in fact) are much more sane in that regard. Dynamic pricing is allowed as a general rule, luxury goods are still overpriced, but essentials like health and housing are regulated to prevent pathological cases like the Epipen thing.

I once implemented SDF text rendering. To me, the thing I liked the most about this technique it how easy it is to add effects on top. With a few lines of shader code, I had outlines and softening of the edges (antialiasing). I didn't try the MSDF variant, as I didn't mind corners not being sharp when scaled up, so I don't know if these effects break on MSDF.

Slug doesn't seem to support any of these, it is just "for a given point, am I in or am I out?", but it doesn't tell you by how much, which is great on really high resolution displays and large sizes, but you would lose the ability to do the kind of effects you can do with SDFs, and have to deal with antialiasing separately.

> If so, why would a private company provide this information?

Because the law says they have to.

Water usage in terms of total volume is meaningless and certainly an overblown concern on a global scale. In fact, it is not a concern at all on a global scale, but it can be a local problem. If a datacenter uses water normally intended for households and the infrastructure is not designed for this, it is definitely a problem. And that's a good reason for having to report water usage, not just to punish them, but also to plan distribution accordingly.

Consumers also have so much data on retailers. It has never been so easy to compare prices and thanks to online stores, availability is at an all time high. Now that eBay and AliExpress exist, I realized how overpriced small items were.

Rental has a problem, made worse by algorithmic pricing, it is, simply, that in some places, housing on a free market would be much more expensive than people can afford. Which is a problem because people need a place to live, and kicking them out is not great. So in this case, government intervention makes sense: a ban on dynamic pricing is one thing, but also just capping rents, subsidized housing, banning short term rentals, etc... the usual "socialist" stuff. But this is, again, not a dynamic pricing problem (though it makes things worse in this case), it is a housing problem.

> they get taken for every penny that they are willing to spend

What is the problem with paying for what you are willing to pay? It is only a problem when you are paying for what you are not willing to pay. Housing is one of these cases, so is food, healthcare and other essentials, also taxes. These are special cases and they usually involve government intervention. And sure, in this case, algorithmic pricing can be a problem, because you leave the usual free market economy.

Price gouging and discounts are exactly the same thing, seen from a different angle. You can look at the people who pay more and complain, or you can look at the people who pay less and give praise.

It is kind of obvious, and supported by economists who are the experts in this field, but the article dismisses it as bullshit without much proof.

It says differential pricing benefits no consumer, and yet, I managed to travel for way bellow cost. If it wasn't for dynamic pricing, I wouldn't have travelled at all. Of course, some people were price gouged, someone has to actually pay for these costs, but these people could afford it, otherwise they wouldn't be in.

The article then mentions overall price increases. Well, yes, sometimes prices increase, for good or bad reasons. Maybe the costs have increases, because there is a war somewhere or something, and the company has no choice but to increase the price to stay in business. Or maybe the company finds itself in a monopoly position and just wants to make more profit. In any case, the price would have increased, dynamic or not. And the solution is not to ban dynamic pricing, it is to avoid getting into wars for the first one, and break down monopolies for the second.

There is also the question of spying on people, but if you don't want spying on people, ban spying on people, dynamic pricing or not. You don't need to spy on people to do dynamic pricing, and many businesses who don't do dynamic pricing spy on their users.

Uber is given as an example, saying that they raised the prices and paid the drivers less. Well, of course they did, at the beginning they operated at a loss, this can't last, at some point they need to make profit. This, by the way, is one of the many shady things Uber has done, a company for which the entire business model is not to play by the rules. The problem is not price gouging now, it is that they were too cheap before, and yes, it is bad, because that's how you unfairly drive off the competition that can keep your prices (dynamic or not) in check later on.

One thing that made me realize the scale of the Apollo project was the "rubber room". This is a bunker connected by a slide to the launch pad. It was designed as an emergency escape in case of an imminent explosion.

At a normal scale, that would be a major construction project.

At Apollo scale, it is an anecdote. Its purpose is to maybe save a few people in the unlikely scenario where an explosion gives enough advance warning to be able to use the slide, but not enough to be able to just drive away. It also goes to show that the program was anything but reckless.

It certainly is, but my mind is more boggled by the Saturn V that did it 60 years ago. Space launch technology feels like it had a 30 year gap, as in, nothing significant happened between the Space Shuttle and the Falcon 9. And now, we are just catching up. Not just SpaceX, there is also Blue Origin, Rocketlab,... and everything that's happening in China.

An by the way, the whole point of Starship is not just doing it, it is doing it cheaply (by space standards). The whole reusability thing is part of it, and so is the "assembly line" approach that lets SpaceX treat their experimental rockets like fireworks. If the goal was just to lift a lot of mass off the ground, it is not even a great design, reusability in particular is costly in terms of payload capacity.

These formulations have been selected by reinforcement learning. People who aligned the LLMs chose this over alternatives.

You know when chatbots ask you which answer you prefer between two. People tend to chose the "as a langage model..." one, so it stuck.

Every time I see a "kids aee getting stupid" article, I wonder what we are missing.

I don't believe that kids are really getting stupid. I see no physiological reasons. Nutrition, pollution, substance abuse,... Not perfect but generally fine. Intellectual stimulation is high, maybe too high. Availability of information is an all time high. If anything the new generation should be smarter than we are.

And yet, it is always "kids these days are stupid". To me, it can't be right. Maybe they are indeed not as good at some tasks. But what are they good at instead? Maybe we should focus on their strengths rather than their weaknesses. We are judging the next generation with criteria of our own, but the world they are going to live in is not the same as ours. Maybe tests should take it into account.

All these fancy weapons are called "force multipliers". And multiplying zero is still zero.

You still need people to occupy the place, take control of objectives, interact with the population, etc... Otherwise, you are just making craters.

An advanced army with appropriate support can make a soldier 10x more effective, but you still need a soldier to begin with.

So ugly...

It looks like a primitive chess engine, trying every move, no matter how stupid, until it works. Relying on its ability to do millions of operations rather than having a plan.

People will try stuff too, but once there is an opening, they will consolidate, generalize, simplify,... before going to the next step. The agents didn't, it is a huge, vaguely directed mess.

Also, it looked so "loud", querying millions of URL with weird requests. The sandbox as weak as it can get, and there is absolutely zero smart extrusion detection or it would have found it. They used their best AI for attacking, but nothing for protection.

I rarely look for positive reviews, these are often the ones that are manipulated. Negative reviews are usually more interesting, they can be fake too, but, marketing campaigns are usually more focused on promoting your own product than undermining your competitors.

You still need judgment, many negative reviews come from people who don't understand the app and where the problem comes from outside the app control. Common with apps that interface with hardware, for example, if you have a remote control app that makes use of a IR blaster, some people will complain that it doesn't work, but that's just because they don't have a IR blaster. But some are clear red flags like "wants me to create an account for not good reason", "intrusive ads", regular mentions of crashes or "since the last version..." comments which suggest enshittification.

About that, a well known problem in VR is that all current headsets have a fixed focal, usually optimized for long distances. The problem is that if you are showing something at a short distance, the eyes will then adjust the focus based on that distance, which is not what the headset shows, so it will look blurry. That's why it is hard to read a book in VR.

Several solutions have been proposed, like varying the focal mechanically based on where you are looking (Varifocal, see Half Dome prototype). There are some others, but the simplest solution is to get old!

With age comes presbyopia, which is the inability for the eye to focus on short distances, bad in real life, great in VR! Because the inability of your eyes and your headset to adjust focus cancel out!

VR is at the intersection of low latency, high throughput, high compute, flexibility and affordable prices.

Embedded systems are often not flexible. They are often designed to do just one thing, like driving an engine, and may run the exact same software for their entire life, which may be decades. They are often limited in term of general computing requirements too.

I don't know HST, but it seems to be similar to embedded: single, highly constrained application, maybe more compute, and I would guess that affordability isn't even on the menu.

Meta VR is consumer software, running on consumer hardware, it runs third party apps, with all the flexibility that is asked of PCs. You can't just put your own realtime OS and claim the entire system for your own specific application.

It is all a compromise. 37 PPD is well under the standard for normal "good" eyesight (aka. "retina" in Apple marketing), which is 60 (aka. 1 arcminute).

Doing more with a narrower field of view is easy, it is even cheaper. Doing more with the same or better field of view means you need more pixels and more computing power to drive them, so more expensive, heavier hardware that gets hotter and with less battery life.

People at Meta consitered 37 PPD to be a good compromise, and I kind of agree.

AMD's random number generator probably can't generate a 0 ;)

Anyways, I think it is good practice to consider the output of any hardware RNG to be biased and to only use it indirectly as a source of entropy. The actual numbers come from a PRNG, secure of not depending on your needs, that have a guaranteed distribution.

It can still cause problems if badly used, or, ironically fix problems if badly used in a different way.

It wasn't a worthwhile skill anyways.

Finding the right words to please the currently popular search engine is not a skill that easily translate elsewhere. It is not like programming where precision is key, searching has always been a bit fuzzy. Fact checking, crossing information, understanding primary sources, etc... these are real skills, typing words in a search engine is meh.

That we now have language models that allow let machines capture the intent of the query without the need for special incantations is a clear win to me.

Where LLMs are really excellent compared to search engines is when you don't really know what you want. You don't need to know about "Cantor diagonal proof", you just need to know "That math thing where you trace a diagonal across decimal numbers", and the LLM will immediately tell you about Cantor.

I think I got the worst sunburns when I didn't feel hot at all.

Usually, it involves water, water cools you down, so you don't feel the heat, but it doesn't block UV, so you don't get the warning.

Of course it could have happened. You don't even need targeting for this to happen. You can drop bombs at random and kill whoever is unlucky enough to be in the wrong place. The US has a long history of doing it (Japan, Vietnam, ...).

So yeah, the military and whoever decided that a war was a good idea are to blame.

People working at Palantir are as much responsible for the killing of these 123 children than seatbelt manufacturers are responsible for the people who died because their product failed. That is, because they didn't do their job correctly, people died, definitely worth feeling bad about. However, should people stop working in seatbelt manufacturing because of it? Absolutely not, we need seatbelts, because without them, things would be worse.

Palantir, or whoever does target identification, is like the seatbelt of the reckless drivers of the militaries, who, for the lack of precise information, would just carpet bomb the place. It is not a hypothetical, it is how the US has worked before they had precision ammunition.

This is essentially a trolley problem. The war is the trolley, people who work for Palantir are the ones pulling levers. And if there is no one to pull the lever, the trolley is gong on a track with a lot of people on it.

It was easily doable for even more years using plain and boring MK1 eyeballs and a stopwatch.

If you are spending millions on ads, you can pay a couple of guys to watch TV and take notes.

I didn't try Zed but VSCode does the same. It is automatic, so you may not realize that it is installing a server, but it does. That it only needs SSH means that it doesn't need an additional communication channel, but you still need to be able to install something. In practice, it means you need a typical non-root shell account.

But Sublime Text can't do that, it is not a client-server application, everything has to run on the same machine, while Zed and VSCode can split the UI part and file access part on two different machines.

The problem with direct democracy is not that people are dumb, it is that they are incompetent.

To do politics right take skills, and I don't expect the average mechanic to be better at it than the average politician is at fixing cars. How should I know if we should subsidize organic farming, ban alcohol sale after 8PM, or increase the defense budget? At least in theory, politicians are professionals who deal with these kinds of questions, they are supposed to know the technical and social implications, or find experts to help them if they don't. Some people think they know, and judging by how stupid most of their ideas are, they don't. I don't blame them, it is just not their field, and my ideas are probably just as stupid anyways.

The answer is probably "no". Sublime Text has no built-in remote editing capabilities.

Of course, it supports remote filesystems (NFS, Windows network shares, etc..), as these are just files, you can also run Sublime Text on a remote desktop or X server, and there as plugins for things like ssh-based synchronization, so its not that you can't do it.

But what it doesn't have and that VSCode and Zed have is the ability to edit files on a remote server with only an ssh connection, by running the GUI part on one machine and the file I/O part on the other.

Even though Sublime Text is my primary editor and I have no intention to change, it is still something I miss. It is one of the reasons I sometimes switch to VSCode (which is a good reminder of how slow VSCode is compared to ST).

If you mean git repos, Sublime Text itself only has minimal git support, but it integrates well with Sublime Merge, which, of course, supports remotes. I don't use Sublime Merge though, it is probably very good, but I never took the time to learn it, I do most of my git work on the command line.

> He destroyed his life and he is a hero for that.

A hero of the world, not a hero of the US. That's what I wanted to say.

Fleeing the country, to a rival country no less is not the best way to come back as a hero. If he really wanted to stand as a hero, he would have stayed in the US and stood for trial, all while being as loud as possible.

It would probably have ruined his life, but being a martyr makes an even stronger point.

Instead he chose to live a (hopefully) good life in Russia, I mean, good for him. The archive is out, he did his thing, he doesn't have anymore duty to anyone. But he pretty much burned his chance to be recognized as a hero in the US. He has his chance in other countries however. He has overwhelming support in countries like Germany and the rest of western Europe, these are nice countries that value privacy and freedom, that's probably where he should go (should he want to and can leave Russia) instead of the US.

That's why I don't really care that much about terms and conditions. Just because they say they can doesn't mean that they will, conversely, the writing can be misleading to anyone who isn't a specialized lawyer, so most people can't really tell what the terms and condition let them do anyways.

I usually go for reputation and precedent rather then legalese. And in the case of Disney, they had ads in their video tapes and DVDs, so I don't need to read the user agreement to know that it has a good chance to come to Disney+, whatever the legal pathway is.

I think it is, again, a way to work around limitations. Economic limitations in this case rather than technical.

Compared to modern HD graphics, pixel art is more affordable because you can hide a lot of details you would normally have to work on behind these fat pixels.

And the most interesting trick, to me, is that pixel art looks sharp! The non-CRT-like square pixels is a feature.

Blurry usually doesn't look good, it is as if your eyes can't focus, or that you are looking through a dirty window, it naturally registers as a problem. Real life has details and you are missing them. With sharp pixel boundaries, you are telling the brain "there is no problem with your eyes, everything is fine".

It is not the only way of introducing artificial details to make low detail pictures look fine: scanline effect, screen door effect, ostensible dithering, etc... But pixel art is a good compromise, in addition to being a recognizable somewhat nostalgic art style.

That's my theory at least.

Except that Microsoft Office is just better.

But with the awesome job Microsoft does at breaking it, maybe in a few years, LibreOffice will be a good alternative, if the LibreOffice guys don't screw up too much on their side.

I actually asked GPT that question some time ago.

The importance of top executive decisions is overstated, they often do no better than random chance. But their decisions are not where their value lies.

The important part is that they own their decisions, or at least appear to do so. And in doing that, they gain trust for the company, they are the company face. It is more of a performance than a technical job. By that, I don't mean it is fake, people will notice if you don't know what you are talking about, and it won't inspire confidence, which is what leaders try to avoid at all costs.

LLMs will probably be at least as good at decision making (I mean, that's almost a dice roll), and they can be convincing, which is important too, but they are completely lacking in the trust department. No one want to work for a machine, no one want to do business negotiations with a machine either. That's why it is important to have a human on top.

There are nuances.

AI scraping for the goal of making a commercial LLM service is different from, say, a commercial file sharing platform.

The first difference is that LLM training is highly transformative. Let's say your LLM ingests the Harry Potter novels during its training. What you get at the other end is not the Harry Potter novels, it is a LLM that can talk to you about Harry Potter, it is not the same thing, and going from one to the other requires a significant amount of work, very expensive work in this case.

Not only that but there is no direct competition. People won't stop buying the Harry Potter novels because a LLM trained on it exists. If you want to read the books, you buy the books, you don't ask a LLM about it. A file sharing service on the other hand competes directly against the official channels, if you want to read the books, you can download it from this service instead of buying it on the official channels.

So, about how free you should be to get these weights from the AI companies. If you just share a 1:1 copy of the weights, that's the "file sharing" situation, not transformative, you took their work, didn't do any of your own. Usually considered unacceptable by IP laws.

Distillation is a more interesting case, you are using a LLM to train your own, it is transformative work, but you may also be competing directly against the LLM you are distilling. So, in a sense it is worse than scraping, but still, despite how much the likes of OpenAI and Anthropic are complaining, it seems to be legal.

So it is somewhat consistent: 1:1 copy and distribution is not allowed, be it source material or LLM weights, and training is, be it source material or another LLM (as in distillation).