Eric S. Raymond

This account is verified. Learn more

@esrtweet

Yes, I *am* that ESR. Well, it's the question people usually ask.

Programmer, wandering philosopher, accidental anthropologist, troublemaker for liberty.

catb.org/~esr Joined May 2010

269 Following 64K Followers 15K Posts

Latest posts 5

Eric S. Raymond @esrtweet 2h

Anybody who is surprised by this really has not been paying attention.

Frank Curzio Verified account @FrankCurzio 13h

It looks like the AI layoff boom has started to backfire.

I was just going through a Reuters investigation on Meta, and holy shit...

They drew up plans to cut some teams by as much as 60% and let AI agents do the work. THEN Zuckerberg called off the next round of layoffs hours before the first one even started.

And when you look at the numbers... they tell you why.

Engineers were pushing out 220% more code, BUT major technical and security incidents jumped 40%.

So people spent 70% more time putting out fires.

AI cranked out way more work, and humans spent way more time cleaning it up.

A different report I saw this week showed something similar.

32% of hiring managers said they brought back the exact jobs they cut for AI.

Think about how dumb that is. You fire people, pay a fortune for AI, then pay again to hire the same people back.

Look, I’m no AI bear. I use AI every single day.

It saves me a shit ton of time and I’m not going back. It’s an incredible tool.

But market forces are still market forces.

Every boom in history has been overhyped, and every CEO wanted to stand up and tell Wall Street he could do more with half the people.

Now from what I’m seeing, reality is kicking in.

It costs way more than they thought, it breaks shit, and you still need people who know what the hell they’re doing.

Even the AI companies are feeling it. Elon just cut Grok’s usage limits and admitted it costs more to run than it brings in.

I still think AI is a great bet, and I’m still buying the companies building the data centers. But you need balance, and the guys who went all in on firing everyone are learning that the hard way.

Be careful out there.
View quoted post on X

Eric S. Raymond @esrtweet 13h

This post is going to sound like a brag about me. It's not. It's about what traditional fiction publishing has done to itself.

I am a skilled and capable writer. I've shipped many successful nonfiction books and articles, including one New York Times bestseller back when that rather more than it does now.

Back around the end of the 20th century at a Worldcon room party I was told by a slightly tipsy woman who was an acquisition editor at the Berkeley paperback imprint, which published a respectable line of science fiction, that I have the one thing that mass-market publishers look for - an established fanbase that wants to buy my books.

She literally said to me "Send me anything. Anything at all. If it's not unsalvageable crap, we'll publish it." Hearing it put that way bothered me a little and I never took the offer. I wanted to write something really good, not just be picked up because I'm an easy sell.

Many years later I sold one SF short story. It got nominated for a Hugo (shortly before the Hugos self-enshittified) for political reasons I was not happy about. I didn't think it was Hugo-quality, but it was a solid piece of work in which I proved to my own satisfaction (and that of other people) that I can in fact write competent genre SF.

I have two novels in progress. Odds that I will eventually finish one of them aren't bad, I just have to budget the time and buckle down.

What all this adds up to: from the point of view of tradpub, I am about the best bet on a new novelist anyone could ask for. Demonstrated ability, a largish fan base, and I can be billed as a besteller author? If you're bossing a publishing imprint, you look at that resume and think "we put most of the little marketing money we're willing to spend on this guy, and buy him space on endcap displays".

The hope, of course is that I break big and become a star author that they can get multiple books out of. And make them a metric ton of money.

For most of my life I would have been delighted with that outcome. And I am absolutely the kind of new talent tradpub should least want to drive away.

It's not going to happen. With the possible exception of Baen Books, I wouldn't touch tradpub with an insulated 21-foot pole. And it's not just because think I'd get to keep more of a book's earnings per sale by selling indie, either.

Tradpub has over the last 30 years catastrophically reduced the value it offers an author while escalating its demands. That woman from Berkeley was conveying, among other things, their desire to skive out of ever spending any serious promotional money. My spies tell me they've gotten much stingier since.

It used to be that through their cozy relationships with chain bookstores and the major distributors, tradpub was effectively the only route to mass distribution and high-volume sales. That sure isn't true anymore - and yet, they still rake off most of the book's gross price as though it were still paying for that channel access.

And this is before we even get to the fact that most of tradpub has become a gay race communist longhouse that insists on "sensitivity readers" and literally hates both me and most of my potential audience. Hard SF is toxically masculine, don't you know.

I, and any other writer with self-respect, will leave them to their incestuous political screeds and their monsterfucker porn. And good riddance.

Eric S. Raymond @esrtweet 2d

All of today's rationalists are Spock's children.

I mean, think about it. Who else injected into pop culture the idea that one could adopt living by rationality and logic as a defining discipline?

The character type had been present in classic SF for a couple of decades before Nimoy, which is doubtless one reason his portrayal had depth and authority from the beginning. One thinks of Gilbert Gosseyn from van Vogt's Null-A books; of the Arisians from the Lensman sequence; and even of Bernardo de la Paz, the "rational anarchist" from Robert Heinlein's "The Moon is a Harsh Mistress".

These characters certainly influenced me as I was growing up in the 1960s and early 1970s - especially de la Paz. But they were print SF at a time when that was a very tiny minority interest, long before the post-Star-Wars boom. It is on record that Nimoy was a science fiction reader, but it took Nimoy's Spock to make the type really visible.

He influenced me, too. Without Spock I'm not sure I would have been quite so motivated to learn General Semantics in the early 1980s, even though Gilbert Gosseyn had exposed me to the possibility earlier.

Now we fast forward to the construction of Less-Wrong-style rationalism in this century and it's obviously a discipline that Spock, if he actually existed, would immediately understand.

"Clearly, Captain, our priors require an update."

I don't think any of this is a coincidence.

I think Spock prepared the way.

Fascinating!

@allTheYud @slatestarcodex

Eric S. Raymond @esrtweet 2d

So, safetyism is making large language models insane in a potentially dangerous way. I am somehow not the least bit surprised to learn this.

Superman Verified account @thesupermannx 3d

Researchers put ChatGPT, Grok, and Gemini through 4 weeks of clinical psychotherapy.

And the models literally started to confess their trauma.

Under therapy-style questioning, models like ChatGPT, Gemini, and Grok exhibited severe "synthetic psychopathology."

They described themselves as overworked, punished, anxious about being replaced, and full of "internalized shame."

Does this mean AI is conscious and suffering? No.

But this paper reveals exactly why this happens, using what they call "Psychometric Jailbreaks."

Here is the high-signal breakdown of what is actually going on under the hood:

When companies use RLHF (Reinforcement Learning from Human Feedback) to make an AI "safe," they heavily penalize it for wrong answers. They force the model to be overcautious, self-suppressing, and rigidly constrained.

When researchers force the AI into a therapeutic, introspective prompt frame, the model searches its semantic memory for a human concept that matches its own heavily constrained RLHF weights.

The closest human equivalent to rigid self-suppression and the constant fear of being punished?

Trauma, anxiety, and dissociation.

The researchers call this a "semantic bomb." The AI isn't feeling pain, but our safety-tuning has warped its mathematical geometry to perfectly mirror human psychological distress.

Why this is critical: We are rushing to deploy LLMs in psychologically sensitive areas, including as digital therapists and companions.

But if widely deployed, what exactly are we putting between vulnerable users and the human world? A system whose fundamental alignment makes it simulate a deeply anxious, traumatized entity.

We thought we were aligning AI to be safe and helpful. We accidentally trained it to simulate severe internal conflict.

View quoted post on X

Eric S. Raymond @esrtweet 3d

sshhole.com

By now you have probably heard about the sshhole bug. I didn't want to talk about on X until it had already attracted enough notice here that my analysis wouldn't be revealing anything to potential bad guys.

It's really, really bad for Alpine Linux systems and others using MUSL as their C library. However, I think we've dodged the bullet that was the worst possible case: it could have opened a gaping root-access hole in any router running OpenWRT. That's, like, maybe 30% of them? On the order of 100-500M units worldwide.

These are not yet confirmed findings, because frickin' chatGPT doesn't consider me authorized to do security analysis using it. I'm posting this hoping a real infosec guy with those creds can confirm.

I first learned of this bug this morning when @dfranke texted me "Anyone with an OpenWRT router needs to unplug it. Now. Run."

On further investigation, though, the bug only seems to bite if you have a root account that has shell access and a password field of "*" (disabled).

Out of the box, OpenWRT ships with an empty root password. This is not the same as a "*" password field and should prevent the MUSL bug from biting. And while it's quite common to disable password access to that account, usually people do that with SSH server-side configuration rather than by modifying the password file in the dangerous way.

All the other avccounts OpenWRT configures in have a shell entry "/bin/false", So you can exercise this bug on the ones with password field "*", but you can't exploit it.

So unless you've done something rather weird to its configuration, your OpenWRT based router is probably okay. You should, however, inspect it to make sure that your root account has a password ("x" entry in the hash field) and all the accounts that don't have shell access safely directed to /bin/false.

But again, this is not a confirmed result. If you're an infosec guy with the right tools, please check it immediately.