Sure thing, AAANUS SNIFFER. It goes a little like this:

1. Step 1: Fire workers.
2. Step 2: Pay out wages of fired workers as executive bonuses.
3. Step 3: Tell remaining employees to make up for the lost productivity or they're fired too.
4. Step 4: Tell investors everything is going great!

Hope that cleared things up for you.

Rust is now 13 years old. When Linus first released the Linux kernel in 1991, the C programming language was 19 years old, and Unix V7 had only been out for 12 years.

There are more than a dozen kernels written entirely in Rust. The claim that Rust is immature is pure propaganda.

I'm not saying he must go; there are alternatives. More like he should be insulated from managerial decisions until he actually has the credentials and social skills that the job requires.

This is a cultural flaw that exists in pretty much every CS department, big and small. There is a default assumption among computer scientists that we do not need bureaucracy or HR. This comes from several sources: the hacker culture around MIT, CMU, and Berkeley was deeply entrenched in the hippie counter-culture of the 1970s; the general absence of women in the field (especially declining during the early decades of the open source movement) somewhat averts one of the most common categories of HR problems; and the paperwork required by external offices and departments tends to be a matter of "following instructions" in a very programming-like way, which goes miles to reinforcing the belief that bureaucracy is merely a kind of programming for non-programmers. (Relatedly, the US banned freelance programming contractors because it feared they would cheat on their taxes, because of a general paranoia about hackers finding loopholes. The actual data suggests programmers have an above-average rate of honesty and obedience when filing.)

So we pretty much universally skip over all the hard-earned lessons from other organizations about the value of managers as diplomats and intermediaries. Fred Brooks made this worse; the Mythical Man-Month demonstrates that traditional corporate structures are inefficient for programming, which we take as validation that we should not wear any yoke that chafes. But it doesn't do or say anything about the baby in that bathwater; it assumes that all programmers are perfectly rational agents with no interpersonal difficulties or competing agendas, working toward the same ends.

The reality is that humans, being humans, are always flawed people, and often need various forms of managing, whether it's encouragement, a sounding board, or policing. Without managers to act as referees, all large open source projects are basically voluntary treadmills: work until you burn out, but only when you feel like it. Some of these duties end up on the shoulders of founders and project leads, but since those people are just programmers themselves, they will invariably be mismatched to the job, and will themselves encounter burnout as they must struggle with the burdens of pretending to be extraverts for the good of the rest of the team.

I don't think codes of conduct are the answer, not really. A code of conduct cannot provide any of the positive benefits of management, nor can it respect the nuances of context. I admit that I think they are well-intentioned, but they're also a half-assed libertarian band-aid for a deep part of the human condition that requires real talent to treat properly.

Creating a toxic workplace tends to get people fired, yes. Next time post from your actual account, you paragon of bravery.

Classy conduct is reserved for classy opponents. Anyone in a de-facto managerial position needs to be capable of functioning as a diplomat. Ted T'so has been pretty fucking clear that he does not have the social skills that are required of someone with his clout. If he were a member of a traditional organization he would have been sent to HR by now.

As for the other developer, perhaps we are thinking of different people. Hector Martin, Asahi Linux project lead, stepped down this week, who actually did cite the "thin blue line" post, along with the general stress of dealing with user expectations. Asahi famously ships another Nvidia GPU driver, called Nova, which is specific to Macs and is written in Rust.

They actually quit over the same post, which was written not by an anonymous "someone else" but by Theodore T'so, a famously shitty person who is the poster child for community codes of conduct. The last Rust-in-Linux dev die-off was also caused by Ted T'so interrupting a live presentation so he could shout about how Rust was a religion and he didn't want to convert, as described in SodaStream's comment. He and his lackeys have, it seems, been repeating lies about Rust (unstable ABI, lack of committed devs) all because he was asked to document how his code works. The narrative he sells about Rust being a flavour-of-the-week thing that he'll personally be stuck maintaining is either delusional or an excuse to maintain his personal job security.

The "meritocracy" is functioning as designed: as a pipeline for enabling tyranny by early adopters.

Unfortunately it seems Linus likes Ted so much that only once the kernel team's name has truly been dragged through the mud will any punitive action be taken.

(Fun fact: deprecate and depreciate are different words.)

The abuse I had in mind was more like the automation of propaganda, SEO, and scamming. Getting an AI to spill censored information or parrot naughty words isn't really an accomplishment as there are plenty of other ways to accomplish these tasks with better results, no matter how much angst it causes for the gilded parasites of Wall Street.

Mea culpa. Yes. Fine tuning, not abliteration. Words.

Before this conversation gets too far into the weeds: the DeepSeek-R1 model itself is actually uncensored. The chatbot on their website has a separate guardrail filter that blocks sensitive discussions, which is why it starts talking but then get cuts off. LLM authors have known for a while that building in censorship using abliteration or another alteration technique damages a model's reasoning ability.

It is worth noting that if the DeepSeek official bot wasn't censored, the CCP would shut it down or block it in China. DeepSeek doesn't really have a choice in this matter, aside from leaving the country.

As open-source LLM enthusiasts are fond of saying, this perfectly illustrates why it's important for LLM weights to not be hidden behind closed doors. Enshittifying AI by putting it behind a paywall may comfort shareholders and convince them that they have a moat, but it does nothing to prevent abuse, since any LLM can be jailbroken with enough skill and patience. (OpenAI used to claim this as a reason for not releasing their models. A thousand shames upon them.)

Unfortunately the only way to actually stop abuse of LLMs is to time travel to 2019 and unalive thousands of machine learning researchers. Pandora's box is open.

(Note: Pandora's box was actually a jar.)

(Amendment: it is.)

Perhaps "berkcoin" was already taken.

Yeah, that falls generally under "free love." I was originally going to put 'polyandry,' which would be closer, but I went with whatever sounded more glib.

You're bang on about nutrition. Testosterone has always been linked to increased height, so a uniform improvement in diet will produce sexually differentiated results. No genetic drift is required to produce these results, although epigenetic markers may cause them to take a generation or two to reach their maximum potential. (Humans retain a handful of nearly-useless epigenetic traits due to variations in chromatin structure. They mainly concern nutritive stress; basically, if your ancestors were constantly hungry, you'll have stunted growth. Plants do the same thing.)

But that said, there is a (somewhat less talked about) genomic explanation for differences in variation between the sexes, too.

We've known for a long time that traits like height, arm-span-to-height-ratio, and intelligence have increased variance in men. These traits are most likely X-linked, i.e. they involve genes on the X chromosome. In developmentally normal women, one of the X chromosomes is chosen at random to inactivate, causing the two to be averaged, which has a strong centralizing effect on the distribution of possible phenotypes. This is not the case for their male counterparts, who get the full force of whatever they inherited from their mothers.

More broadly, sexual selection in humans, and indeed nearly all mammals, treats males as a sort of genetic lottery or test kitchen for new mutations and gene combinations. If they "win," so to speak, then they find a mate, and half their DNA is folded back into the tribe's treasury. But they're never not playing the game: 52% of newborn humans are male. While the 4% surplus obviously buffers against losses during hunting accidents, it (perhaps more importantly) encourages competition by ensuring that there are, eventually, not enough females to go around, thereby denying the least-competitive males the chance to pass on their genes. (Alternative theories include "human nature favors free love," "human nature favors rape," and the modern classic, "human nature favors the emergence of incels." Pick your favorite explanation!)