Tests are as important as the code they test. It's just as essential that they be correct. If you're going to have AI write your tests, you need to review them, understand them, and verify that they really test what you intended. Otherwise you're likely to get a lot of useless tests that report success even when the code is wrong.

I'll sometimes intentionally break a piece of code just so I can see the test fail. I need to prove to myself that the test really does fail if the thing it's testing is wrong.

Try saying that when you have a big code base with thousands of tests, and it takes an hour to run them on every commit. Efficiency is very important for tests.

Once the EV market share reaches a certain level, it tends to explode. You see the same pattern repeated across a lot of very different countries.

Norway led the world. Their EV market share hit 15% in 2014. Three years later, it was up to 39%.

Iceland was a few years behind. They reached 14% in 2017 and three years after that were at 52%.

China followed exactly the same pattern. They reached 15% in 2021, and three years later had reached 48%.

Vietnam grew even faster. They were only at 10% in 2023, but had already reached 40% in 2025.

Once a country reaches about 10-15% market share for EVs, it tends to explode in just a few years. In this list of 55 countries, 31 have EV market share of at least 15%, and 10 more have at least 10%. Expect most of those countries to pass 50% within just a few years.

Isn't it great how they always say the problem has already been fixed? It's a requirement for any time they've done something wrong. They always assure us the problem has already been fixed, so we don't have to worry about it happening again. Yet somehow the problems keep on happening anyway, and in fact are rapidly getting worse.

It almost makes you think they're lying when they claim every time to have fixed the problem. More like they keep rushing out band-aids they can slap on to pretend they're addressing the problem, without actually addressing the core issues that make these problems inevitable.

When conditions change slowly, life adapts and thrives. When conditions change quickly, it doesn't have time to adapt and just dies.

We are currently adding CO2 to the atmosphere ten times faster than during the End Permian Extinction, which came frighteningly close to destroying all complex life on Earth.

These satellites are intended to go into an orbit that passes overheat only at sunrise and sunset (the dawn/dusk orbit mentioned in the article). That's the only orbit that puts them in full sun all the time. Any other orbit would leave them in darkness half the time, which would eliminate the claimed advantage of putting them in orbit in the first place (solar energy 24 hours/day).

That means all these satellites will be spread out in a single ring around the earth, and that ring will pass overhead twice each day.

Nearly all countries have some level of restrictions on possessing guns, drugs, explosives, and various other things considered harmful. There also are many things that it's legal to possess, but only if they have certain safety features to prevent them from hurting you or other people. So what exactly is the position you're expressing?

- All those restrictions are bad, and none of them should exist?

- Restrictions on some things are ok, but AI shouldn't have any restrictions at all?

- Some restrictions on AI might be ok, but they'd create a slippery slope and lead to worse restrictions, and therefore we should oppose even reasonable restrictions?

- Something else?

Actually, scratch what I just said because the situation is much worse. A sun synchronous satellite only passes over at sunrise and sunset. If it's noon where you are, the nearest satellite is 90 degrees of longitude away from you. Sending your jobs to a nearby satellite just isn't an option.

The FSF position seems kind of misguided to me. The way you maintain the basic freedoms they're talking about is to use open weights models running on hardware you control. If you delegate to a company running their own models on their own hardware, that freedom is already gone. Someone else is in control, not you.

At that point, giving the government an extra point of control doesn't change the basic situation and could be a good thing. That's not to say I trust the government, but I trust OpenAI even less. I'd like to think someone who isn't guided entirely by profit has the ability to step in and stop them from doing dangerous things.

Yes, but that's only useful for short jobs. A satellite that's overhead now will be on the other side of the planet in about 45 minutes. Anything that runs for a long time has to deal with high latency and breaks in connectivity.

And if you only send jobs to nearby satellites, that creates other problems. The earth's surface is 2/3 ocean, and large parts of the rest are mostly uninhabited (Antarctica, Siberia, etc.). Your satellites will spend most of their time over uninhabited places, meaning they spend most of their time idle and you need several times more satellites to make up for it.

It's more complicated than that. The plan is to put them in sun synchronous orbits so the solar panels always have light on them. Unfortunately, that's an orbit that passes over every point on earth.

This isn't like a Starlink satellite that just acts as a network connection. For those, you can use whatever satellite happens to be passing overhead. An orbital data center satellite is unique. If you start a calculation running on one satellite, you have to keep communicating with that satellite. When it's on the other side of the earth, you have to communicate with the other side of the earth. When it's over the middle of the ocean, you have to figure out how to communicate with a satellite over the middle of the ocean.

For some applications, this doesn't matter much. For short calculations that complete in a few seconds, you just pick a satellite that's nearby. For long running things, it can be a big problem.

Here's the story from just over a month ago when they put out their report on the Hugging Face breach. This is what they said at the time.

Got that? It was "an outlier scenario involving a rare and unexpected confluence of events". Not something you can expect to happen often. They did everything they could to downplay it. Now here we are, a month later, and they've found over 100 similar incidents, with the review still ongoing and expected to take months more to complete.

So we can completely trust that they're not still trying to downplay it and it won't continue to get worse.

"Hot electrons" doesn't mean the panel is hot. In fact, that's the point of this research. They found a way to slow the transfer of energy from the electrons to the crystal lattice. The longer you can keep it in the electrons, the more opportunity you have to convert it to electricity instead of heat.

If you just want to convert sunlight to heat, that's easy. A piece of black plastic works great. That's how solar water heaters work, and they already have efficiency way above 33%. Converting any kind of energy to heat is easy, because that's where it's going eventually anyway. Preventing it from turning into heat is what's hard.

The Linux kernel is not from GNU. They had nothing to do with it.

An increasing number of linux distros use musl instead of glibc and llvm instead of gcc.

"Majority of the userland"??? You mean other than the display manager, windowing environment, package manager, init system, file browser, web browser, media player, office suite, email client, and almost everything else that people run in userland? (And I'm just talking about the bundled applications that come preinstalled.)

I predict they'll go right on saying it. It's been many years since GNU code made up more than a tiny sliver of the code in most Linux distros. That didn't stop them. I doubt this will either.

Are you thinking of the 2021 Texas winter storm, when renewables held up great but production from gas and coal collapsed? Of course, the governor then lied about it and tried to claim the opposite: