Yes!

https://en.wikipedia.org/wiki/Lunar_Flag_Assembly

However, the US is a member of the Outer Space Treaty, so it can’t claim celestial bodies.

https://en.wikipedia.org/wiki/Outer_Space_Treaty

Article II of the treaty explicitly forbids any government from “appropriating” a celestial body such as the Moon or a planet, whether by declaration, use, occupation, or “any other means”.

I think that SDL can do aalib and libcaca.

searches

Hmm. No. I must be thinking of mpv.

SDL 1.2 could do aalib, but apparently 2 cannot:

https://wiki.libsdl.org/SDL2/FAQUsingSDL

If there’s the demand, and I bet that there will be, someone is going to be making plug-and-play local AI compute boxes once prices come down that eliminate a lot of the learning curve. Like, preconfigured software and all that.

Joe User is not going to be talking about memory bandwidth or compiling software or setting sampler parameters. But—as the cloud guys demonstrate—he can use it if it’s set up. And there will be more-approachable front ends and such.

Use a local model

I also think that local models are generally preferable, but everyone having a local model requires more hardware, since much of the time the local hardware to run the local model is probably going to be idle, whereas hardware in the cloud can be servicing User B when it’s not servicing User A.

In 2026 and 2027, absent some very unexpected development, memory prices are going to be elevated.

And while some people could get local hardware, we couldn’t build a comparable-to-what-someone-could-get-in-the-cloud AI compute box for everyone if everyone wanted to go local, not for years, as we don’t have the memory production capacity today. If everyone tried to get a local box for AI compute at the same time, it’d just do what happened when the cloud companies did it, but at greater scale, because it’d require even more memory. There’d be a new memory shortage. Prices on memory would keep rising, pricing would-be buyers out, until the number of buyers still willing to buy was equal to what supply was present.

Right, in cases where an acronym is pronounced as word it generally follows the normal convention of the letters that form the word.

Meanwhile, the GNU Project calls itself “guh-nu” and then chooses a gnu (“nu”) as its mascot.

Technically the original JPEG container format is “JIF”, throwing another wildcard into the mix.

https://en.wikipedia.org/wiki/JPEG_File_Interchange_Format

The base specifications for a JPEG container format are defined in Annex B of the JPEG standard, known as JPEG Interchange Format (JIF).

Who the hell says “Jif?” Jiff is a god damned penut butter.

The creators of Jif call it Jif, and they’re clearly authoritative!

If someone is concerned about climate policy and wants people to curtail their ICE vehicle use, high fuel prices are probably going to be the only realistic way that this is gonna be exposed to the consumer (like, we probably aren’t going to have an administration impose fuel rationing or something) so this is pretty good. Not only are they high, but it’s also good odds that your favorite politician isn’t the one taking the political hit for it. As the article points out:

In the U.S., gas prices were less than $3 per gallon before the start of the Iran war in February, and since then, the national average is now $4.14. That increase has cost Americans more than $50 billion, according to a Brown University analysis. As of this writing, the most expensive gas in America is in California, where car owners are spending as much as $5.78 per gallon. That has hurt the economy, with Americans spending a lot less money on vacations over the summer. Flights and road trips have also declined, as jet fuel costs have also been affected by the Iran war.

https://www.technologyreview.com/2026/03/26/1134677/gas-prices-electric-vehicles/

Historically, this is exactly the sort of moment that’s pushed people to reevaluate how they get around. During the oil crisis of the 1970s, Americans switched to smaller, more efficient cars in droves. It was a major opportunity for Japanese automakers, whose vehicles tended to fit this mold better than those produced by their US counterparts.

Nice, round numbers do tend to get people’s attention. Some point to $4 per gallon (which the national average is quite close to right now). At that price, the total cost of ownership for an EV is comfortably lower than the cost for a gas-powered car, even with higher electricity prices, according to data from the energy consultancy BloombergNEF.

Then again, maybe that won’t quite do the trick: One survey from Cox Automotive found that most US consumers would consider switching to an EV or hybrid if gas prices hit $6 per gallon.

Mind, I don’t think that Trump is gonna get a Greenpeace or Sierra Club endorsement any time soon…

The alternative, Vance said, was accepting that “the Iranians were going to shoot at ships like crazy people.”

Vance, the Trump administration was shooting at Cuban boats to score political points less than a year ago.

deleted by creator

It actually is. I mean, it’s building a dictionary off of a variety of content ahead-of-time, rather than training it on the specific item in question, but that’s not uncommon for non-general-purpose compressors.

I mean, doing so to a (probably short) prompt is (a) lossy (and I gave a lossless example) and (b) lossy to an extreme degree, to where it’s probably not incredibly useful option for the kinds of systems that exist today.

But…existing diffusion models aren’t actually intended for this, either. I’d bet that you could train a model to do image compression along these lines, with a large dictionary, that could do usable compression along the lines of what is (jokingly) described in the article. Probably have a larger compressed form than what they’re thinking of.

EDIT: At one point in time, about over a quarter-century ago now, I went out and banged on a neural net post-processor for JPEG artifacts. The idea here is that JPEG very probably isn’t optimally representing the final image, as a human, using their knowledge of what the world looks like, can manually (if time-consumingly) clean these up. I didn’t meet with a lot of success; I only wanted to put a small amount of time into it, and I was working with much weaker hardware than people are running neural nets on today. But that generated a pre-existing dictionary, a pre-trained neural net, off a training corpus of uncompressed images. It didn’t try to reconstruct the image from scratch, the way something like this would, just clean up artifacts, but it has that same pre-generated neural net approach.

Despite the fact that it’s registered and paid for until 2040.

I can believe that Verisign might have some route to terminate service (even in cases where there is some contractual obligation to provide a service for a given period of time, someone can typically break the contract, just having to pay some sort of penalty for doing so), but I also suspect that they are legally required to issue you some sort of refund.

You generally cannot just sell someone a service and then refuse to provide it; that’ll violate consumer protection laws.

That probably won’t make the author happy—he probably wants the domain, not the money. But at the minimum, I will say that he probably doesn’t have to pay for service that he doesn’t get.

It might be that they are doing that; he doesn’t specify.

But it gets much worse. Once the 3rd-level domains are terminated, it is assumed that the now vacant 2nd-level domains will become available for registration. Should someone (other than me) scoop up fraser.name they would be able to recreate and control neil.fraser.name. They’d be able to hijack hundreds of accounts that are linked to that address. They could commit code with my authentication. They could seize control of IoT devices. There is no way to enumerate all accounts (online and offline) which have been opened using this email address over the past quarter century.

Yeah, personally I’d rather have some kind of key-based mechanism, at least as an alternative to the X.509-based system. Think of what PGP or SSH do. Like, the key is your identifier, not the endorsement of some centralized registrar.

Hyphanet does something like this, can do key-based addresses.

That has some drawbacks too—I mean, you are the ultimate authority here, and thus responsible for keeping that key safe, and that’s a big responsibility. If you think of all the attacks on cryptocoin wallets and stuff, I think that personal computers are not incredibly-well-suited for that—if we’re going to do that, we probably want something more like hardware-based keystores, like YubiKeys or similar, where a key cannot be exfiltrated if a PC is compromised, as PCs have a big attack surface. And it’s harder for humans to remember a key as an identifier, which means that they probably have to have a computer somewhere store a list of known and trusted keys (though doing so solves a lot of existing attacks, like phishing emails with look-alike URLs that attack the limited human ability to remember names, so there are pros and cons). And it avoids the risk that someone with access to a certificate authority’s key can compromise your security.

There is a “fighting words” exception to the First Amendment, which has been narrowed by courts so far as honor culture has declined that it’s mostly non-operative today. That isn’t a defense for someone who attacks someone who says something, but it did recognize that there was a social culture at one point of not being able to just outright insult someone without them having a social obligation to respond violently, so it permitted laws against such insults.

https://en.wikipedia.org/wiki/Fighting_words

Fighting words are spoken words intended to provoke a retaliatory act of violence against the speaker. In United States constitutional law, the term describes words that inflict injury or would tend to incite an immediate breach of the peace.

The fighting words doctrine, in United States constitutional law, is a limitation to freedom of speech as protected by the First Amendment to the United States Constitution.

In 1942, the U.S. Supreme Court established the doctrine by a 9–0 decision in Chaplinsky v. New Hampshire.[1] It held that “insulting or ‘fighting words’, those that by their very utterance inflict injury or tend to incite an immediate breach of the peace” are among the “well-defined and narrowly limited classes of speech the prevention and punishment of [which] … have never been thought to raise any constitutional problem.”

Chaplinsky decision

Walter Chaplinsky, a Jehovah’s Witness, had purportedly told a New Hampshire town marshal who was attempting to prevent him from preaching that he was “a God damned racketeer” and “a damned fascist” and was arrested. The court upheld the arrest and wrote in its decision that:

There are certain well-defined and narrowly limited classes of speech, the prevention and punishment of which have never been thought to raise any Constitutional problem. These include the lewd and obscene, the profane, the libelous, and the insulting or “fighting” words – those which by their very utterance inflict injury or tend to incite an immediate breach of the peace. It has been well observed that such utterances are no essential part of any exposition of ideas, and are of such slight social value as a step to truth that any benefit that may be derived from them is clearly outweighed by the social interest in order and morality.

— Chaplinsky v. New Hampshire, 1942[1]

The idea here is that you can’t just go around calling someone a fascist without expecting them to immediately shoot you or otherwise hurt you in response, so to avoid that, it’s constitutionally-acceptable for legislators to pass laws to prohibit someone from calling someone else a fascist like that.

Over the years, it’s been steadily narrowed by a series of court decisions.

Post-Chaplinsky

The Court has continued to uphold the doctrine but also steadily narrowed the grounds on which fighting words are held to apply. In Street v. New York (1969), the Court overturned a statute prohibiting flag-burning and verbally abusing the flag, holding that mere offensiveness does not qualify as “fighting words”. In similar manner, in Cohen v. California (1971), Paul Robert Cohen’s wearing a jacket that said “fuck the draft” did not constitute uttering fighting words since there had been no “personally abusive epithets”.

In Brandenburg v. Ohio (1969), even speech such as “Bury the niggers” and “Send the Jews back to Israel,” was held to be protected speech under the First Amendment in a per curiam decision. In addition, despite the speech being broadcast on network television it did not direct to incite or produce imminent lawless action nor was it likely to produce such action.

In 1972, the Court held that offensive and insulting language, even when directed at specific individuals, is not fighting words:

• Gooding v. Wilson (1972): “White son of a bitch, I’ll kill you.”[2]
• Rosenfeld v. New Jersey (1972): “mother fucking.”[3]
• Lewis v. City of New Orleans (1972): “god damn mother fucking police.”[4]
• Brown v. Oklahoma (1972): “mother fucking fascist”, “black mother fucking pig”. Found constitutional because the “speech [may] have been anticipated by the audience.”[5]

In Collin v. Smith (1978) Nazis displaying swastikas and wearing military-style uniforms marching through a community with a large Jewish population, including survivors of German concentration camps, were not using fighting words.

Texas v. Johnson (1989) redefined the scope of fighting words to “a direct personal insult or an invitation to exchange fisticuffs” in juxtapose to flag burning as symbolic speech.[6]

In R.A.V. v. City of St. Paul (1992) and Virginia v. Black (2003), the Court held that cross burning is not ‘fighting words’ without intent to intimidate.

In Snyder v. Phelps (2011), respondents’ counsel argued that the Court’s definition of fighting words required immediacy, imminence, intent and proximity. Justice Ginsburg stated that the Court had rejected spreading the concept beyond words that immediately trigger an instinctive reaction.[7] The Court held that even “outrageous” and “hurtful speech” such as: “God Hates the USA/Thank God for 9/11”, “America is Doomed”, “Don’t Pray for the USA”, “Thank God for IEDs”, “Thank God for Dead Soldiers”, “Pope in Hell”, “Priests Rape Boys”, “God Hates Fags”, “Fags Doom Nations”, “You’re Going to Hell,” and “God Hates You” is to be considered public debate, particularly when conducted on public land, and must enjoy “special” First Amendment protection. Lone dissenting Justice Samuel Alito likened the protests of the Westboro Baptist Church members to fighting words and of a personal character, and thus not protected speech. The majority disagreed and stated that the protesters’ speech was not personal but public, and that local laws which can shield funeral attendees from protesters are adequate for protecting those in times of emotional distress.

It’s humorous, but last I checked, the best general-purpose compressors with the highest levels of compression—even lossless, which is probably not what most people think of when they think of neural nets—are neural net based.

Neural net-based compressors are computationally expensive, which is why we don’t normally use them for most day-to-day tasks, but they really can produce really small outputs.

I’m going to take the text of the US Constitution and stick it in a text file.

$ wget https://www.gutenberg.org/cache/epub/5/pg5.txt
$ stat -c %s pg5.txt 
48326

Okay, so 48326 bytes.

Let’s do lzo. You’d expect a limited amount of compression — LZO is “fast” compression, usually only used where compression speed is really important, like where you want to be compressing stuff that’s going to be decompressed once and your bottleneck is throughput to disk:

$ lzop <pg5.txt >pg5.txt.lzo 
$ stat -c %s pg5.txt.lzo
24843

Okay, how about gzip? That’s Deflate, an older, but pretty-widely-used general-purpose compression algorithm.

$ gzip <pg5.txt >pg5.txt.gz
$ stat -c %s pg5.txt.gz
16660

Okay, what about LZMA? That’s a newer, more-CPU-intensive thing that’s probably a good general-purpose choice that’ll generally give better compression ratios. It’s the kind of thing that I’d probably use in a lot of cases. (Personally, these days, I tend to use pixz, which provides both indexed access for tarballs and parallel compression and decompression, which is important for modern processors.)

$ xz <pg5.txt >pg5.txt.xz
$ stat -c %s pg5.txt.xz 
15488

Okay, now PAQ, a neural-net-based compressor:

$ zpaq a pg5.txt.zpaq a pg5.txt -method 5
$ stat -c %s pg5.txt.zpaq 
13063

A witness, who identified himself as a cousin to both men, told investigators that the group had spent the afternoon drinking alcohol and cooking green chile enchiladas.

Alcohol and enchiladas: a recipe for irresolvable conflict.

The dispute escalated outside the apartment, leading to a physical fight in which Cavazos punched Gutierrez, leaving him bleeding, the witness reported.

After the fistfight ended and Cavazos began walking away, Gutierrez allegedly drew a handgun and confronted him, shouting, “We’re family! We’re family! How could you do this to me?” before opening fire, according to the witness statement in the affidavit.

Cavazos was shot three to four times in the back from roughly 10 feet away while unarmed, the document states. Gutierrez then fled the area.

“Family would never beat someone else bloody!”

Promptly proceeds to shoot family member in the back.

aww goddammit what did RFK jr. come up with now?

This is apparently a Hegseth thing.

Yeah, I’m very much Team Lone Monitor. You can only see sufficient detail to work with text in a limited area in front of you. I can flip between desktops in sway faster than I can look from side to side. Honestly, my current monitor is already probably excessive in screen space for productivity stuff.

And the ergonomics are preferable if you’re using secondary monitors much. You ideally want the display in front of you:

https://worksmarthub.com/monitor-ergonomics/

If you only change four things today:

Adjust your chair height first.

Raise your monitor so the top is at or just below eye level.

Position it about an arm’s length away.

Centre it directly in front of you.

These principles align with ergonomic research on neutral posture and reduced upper‑limb load (Rempel et al., 2007).

I remember working on a dual monitor setup split in the middle back in the 1990s, and while it was never a huge issue, after a day of working, my neck was stiff from looking to the left.

Added bonus of a single-monitor setup: less power use, and you can work on a (more-portable) laptop without extra screens without breaking your workflow.

EDIT: Note that this is just talking productivity stuff. If you also play games or something on the machine, then having stuff in your peripheral vision can have value, just as in real life. VR rigs push that even further.

EDIT2: If someone does come up with some sort of software that makes useful use of peripheral vision for office stuff, someone will be selling VR products aimed specifically at this in short order. I remember in…the early 2000s maybe, when Intel was trying to convince businesses to buy 3D hardware, and had ads talking about how 3D chart viewing would increase productivity.

https://en.wikipedia.org/wiki/Linus_Van_Pelt

Well…the in-compiler time analysis I think wasn’t as good then, though I recall doing my own ad-hoc profiling, logging time for each unit. But the refactoring stuff should be available.

I dunno how far back distcc goes.

checks git repository

The git log goes back to 2008, so almost 20 years, but it’s also talking about release 3.0 then, so there might have been time outside a VCS or with history that didn’t get imported.

Wikipedia says that ccache goes back 24 years.

I mean, none of the optimizations I listed rely on multiple cores.

It’s true that just running parallel compilation jobs on your build machine is a simple way to speed things up, no disagreement, but it’s not like optimizing compilation time isn’t a thing outside of that.