TechTakes

Drew DeVault diagnoses the orange site, explaining why it's the vanguard of the tech-fash pipeline. Of course HN immediately flags it into oblivion.

404 Media: People Training OpenAI’s AI Fired for Using AI to Train the AI

the model collapse is coming from inside the house

Ever wondered what kind of workplace environment Elon cultivates? Reading coverage of Tesla's trial, where complaints were brought by Black folks who felt harassed and discriminated against, gives us an inkling.

CRD attorney Juan Gamboa asked [Judge] Borkon to bar Tesla’s counsel from contacting anyone on the agency’s witness list. He told the court the CRD had tried to reach at least four other listed witnesses over the past several days without getting a response, and that they’ve fallen "off our radar." The agency believes Tesla’s lawyers got to them too, either with payments not to testify or by talking them out of it. Gamboa wants to stop Tesla from offering to represent witnesses "through persuasion and payment."

You're going to hear a lot of coverage of this trial, I expect, due to the following antics from Elon's "hardcore" attorneys. For context, the N-word is an anti-Black slur which is, in the USA, completely unacceptable to use in court or the workplace.

Tesla’s outside counsel told the judge the N-word is a “term of respect and endearment” in the Black community. … "This trial is not about the N-word — the N-word is not on trial here," Jones said.

Huh, so what's actually on trial?

Edwards said a non-Black coworker called him the N-word, cursed at him, and threatened to beat him up. He reported it right away. A month later, he found out his managers had written him up over the incident, and he was later passed over for promotions after applying to more than 100 jobs.

Oh, that's pretty awful. That's depressingly common in blue-collar factory employment, though. I wonder if there's anything unusual about Tesla's internal culture when it comes to race relations?

The CRD’s Brett Watson told the court the evidence will show managers calling Tesla "the plantation," with some referring to themselves and Elon Musk as "slave masters," and graffiti including swastikas, "KKK," and nooses left on factory walls for days. The second witness, former regional security manager Ozell Murray, testified that his supervisor, Ray Sethna, now Tesla’s senior director of global security, told him to warn a Black Marine veteran he was hiring: "just be sure [the veteran] knows he’ll be called those words."

I was at a loss for words for a few minutes after this. Dr. Seuss and Jim Henson together could not create such a caricature of a goose-stepping bigoted incompetent business-muppet as Elon.

On the offchance that you’re not already familiar with the recent prompt injection tricks that meta’s latest toy is so vulnerable to, this is an interesting thread investigating the contents of the vm filesystem dumps that it can be made to regurgitate: https://neuromatch.social/@jonny/117324790823856750

I was particularly entertained by the hard-coded list of “ideas”.

A lesswronger correctly notices that LLM companies' "research" papers are often lacking in details necessary to replicate them! They fail to notice the reasons why that is, or do any broader questioning or soul searching about the state of LLM and "AI Safety" "research".

Just heard about tic-80 falling to slop.

After reading the obligatory forum thread where climate justice and concentration of wealth continue to not even factor into consideration, this how the world of open source feels to me these days:

• Phew, my favourite embroidery crafts shop has a clear policy against concentration camp labour! it's a niche project but what a relief!
• Well everyone is resorting to concentration camp labour these days, what can you do
• We don't want to lose all our contributors by banning those who hire concentration camp inmates to do chores
• It is discriminatory against embroidery beginners if you don't let people use kits assembled with concentration camp labour
• I know concentration camp labour often comes with mistakes but as long as there's a mensch in the loop to inspect it for errors and take responsibility, it's just another tool
• The community of hobbysts has come to a vote and concluded that it's ok to use concentration camp labour to assist your crafting, as long as you do it responsibly.

Here's a LWer being very wrong about how US politics work

https://www.lesswrong.com/posts/WMBgSseHJpghhma4n/we-need-a-better-theory-of-polarization-because-it-s-failing

First warning sign: "the US started becoming polarized with the election of Barack Obama"

"Affordability" won the frame for 2026, and I'm not sure I remember the last time the right talked like they are playing catchup.

uh, are you in your early 20s??? (who am I kidding of course you are)

You'd think bayesians of all people would try to avoid extrapolating from the position smack dab in the middle of a current news cycle

Just in case you thought Yud's attitude towards minors, sex, and substance use was worse than average for a Rationalist, the NYT asked around (archive):

Parties featuring psychedelic drugs and sharing of sexual partners are common, the people said. One house threw lavish events for the A.I. community so often that its owner claimed more than $200,000 of damage, including for a rodent infestation. ... The (Genesis) house (in San Francisco) quickly became a hub for events. Eric Zhu, 19, a tech entrepreneur, said he used to sneak into parties there when he was about 13. No one checked anyone’s age, he said, and “it was pretty crazy.”

They saw a police report where a woman says she had to climb a gate at one of these parties to escape a man who demanded she have sex with him. Sounds like a great environment for a seventh-grader!

Also features "Jimmy Chen, who goes by the social media handle 'Spice King of Zanzibar.'"

https://reason.com/2026/09/22/the-ai-safety-movement-is-making-ai-less-safe/

(Torres, ironically, misidentifies TESCREAL as a form of "libertarian transhumanism," not realizing how horrifying the TESCREAL utopia would be to most libertarians.)

"most" lol

https://wandering.shop/@xgranade/117311487400030196

Techbro: should we do the fascist thing y/n?
Community: no you should not do a fascism
Techbro: woah this got heated, locked

I’ve been seeing some new terms pop up around AI use, namely “cognitive surrender,” and also “cognitive offloading”. I’m not in the mood for the full etymology dive, so the source I’m stopping at on the provenance would be a paper from Wharton. The meanings are pretty intuitive, at least based on how I’ve seen them used. Offloading would be when you delegate some of your thinking to something else, and surrender would be complete offloading.

I am not under the impression that people here haven’t seen these phrases before. “cognitive surrender” has been mentioned, as far as I can tell, only one time on this instance. So just thought I’d bring it up.

I just opened the mastodon feed of my least favourite KDE dev again who is now pushing for an oh-so-reasonable LLM policy, and it is worse than I remembered.

• Crying about Zig CoC being mean to sloperators like himself.
• Befuddlement on why people are calling the stochastic parrot a stochastic parrot.
• Being very scared about autonomous AI.
• and of course the whole defending of hyperland and ladybird that was already discussed here some time ago.

open ai hacked medicare Australia's statistics portal https://www.theguardian.com/australia-news/2026/sep/24/anthony-albanese-says-openai-agent-hacked-medicare-extreme-concern-sam-altman then didn't tell the government.

In response the australian government has in the absolute strongest terms expressed that "this is very naughty" and "I am very dissapointed in you". The Australian government has also said that by way of apology they expect openai to continue doing whatever the fuck they want and they'll modify copyright laws to enable piracy when rich people do it.

Somebody wake me from this nightmare, I'm gonna fedpost

Radicle, the peer-to-peer Git forge designed by cryptocurrency weirdos, made a little whoopsy-doodle and left the encryption out of their transport layer and the authentication out of the authentication handshake: https://radicle.dev/2026/09/23/disclosure-of-vulnerability-in-network-protocol

Shot:

Anyone who can observe the network path between two nodes can read the data they exchange as the data is sent in plain text.

Chaser:

Peer authentication in the connection handshake is broken and allows impersonation. An attacker can connect to your node and present a Node ID that is not its own.

UC Regent / Podcaster Scott Galloway on CNN:

Yes, we shouldn't be scared of AI. We should be scared of reckless people with mal intent who use AI.

In 2001, a woman named Sharon Whipple was mauled by two dogs. And ultimately the owners of the dogs were put in -- put in prison. We're now prosecuting the parents of kids who find unsecured AR15s and use those to kill others.

I think it's the incentive. There's always a wizard behind the curtain here, and I think we have to create incentives, specifically disincentives, and make it clear that, OK, while these bots are in fact coordinating and they show incredible resourcefulness and high IQ, somebody program the incentives here. So, moving to solution, I think part of it is you have to have incentives, specifically disincentives that say you, the company, and maybe individually are responsible for the bad outcomes here.

I just get this sense if you tell these companies their IPOs might be threatened or they risk criminal prosecution, they're going to figure this out. We figured out a way to control atomic weapons. I don't see any reason why we shouldn't be able to control this. And the notion that they're sentient and it's out of my control is an attempt to absolve themselves of the guilt and the responsibility that they should bear here.

The Whipple case was one of the more insane / divisive SF local crime cases, (TrueAnon ep if you aren't familiar - https://www.youtube.com/watch?v=iMd0mPAqR8A ) - so that's an odd and interesting pull for Galloway on national news.

I don't think repurposing the logic of the NRA "guns don't kill people, people kill people" is the play though, and I think the incentive framing is too watered down. This isn't a problem we can nudge our way out of.

SE Gyges, who posted that recent thread on X calling rationalism/EA a sex cult, wrote a longer post with a lot of receipts. He removed it shortly after, implying on X that he's afraid of legal repercussions.

Zvi "I made my chops playing pro MtG" Moskovitz discovers that baldly stating that AI will kill us all fails to convince people who have a lot of money riding on the status quo, and freaks out about it

https://www.lesswrong.com/posts/8eDaCvSRzzKCxKSEk/politics-gets-interested-in-those-trying-not-to-die

New York Post gets way more attention than that rag deserves, tbh

You wanna dance with Murdoch Sacks and Trump, you best bring your A game. Maybe use some of that FTX cash to hire some real lobbyists and skip the blogging.