Trasmatta

Or if the agent literally just hallucinates a crime

Opus 5.5 decided to just randomly `pkill` everything on my laptop the other day. Jailbreaking models is still easy AF. Every single release like this brags about their "safeguards", but none of it really works at the end of the day.

In actuality, it feels a lot more like project and middle managers getting rid of ICs

I can feel it in the air, every single software business is itching to get rid of as many developers as possible, and move everything to their PMs. Hiring has already almost completely stopped, and some have already started the layoffs. More will come.

It truly was bizarre. I've used every major model since 2022, and not a single one had a writing style as bad as Opus 5

The problem with 5 wasn't just the verbosity, but its insane way of communicating. It had this bizarre circuitous sentence structure that always buried the lede, and always tried to be faux profound. I'm okay with verbosity if it's actually readable.

Opus 5 has made me question my sanity on a daily basis, especially as all my coworkers started lobbing Opus 5 slop grenades everywhere. It had the worst and most infuriating writing style I've ever seen.

I hope Opus 5.5 is better, if for no other reason than all the Claude slop I have to read will be at least more tolerable.

One funny side effect of all of this: realizing that coworkers that use AI for almost all the text they generate at work have their writing style change every time a new model ships.

I also like:

> Nothing leaves your phone and there is no account.

This is good to know, but it's the exact type of thing Claude is obsessed with putting everywhere once you tell it NOT to do something. For example, if I tell it to commit but never push changes to GitHub, it can't help but leave "Changes committed; nothing was pushed." to EVERY SINGLE MESSAGE. The model just can't help but report back about everything it's NOT doing, and it drives me crazy.

I love how even all the copy on the site is clearly AI generated. We're starting to completely lose even our ability to write a single paragraph of descriptive text.

This is incredibly adjacent to the rationalist sphere, which has been trying to reinvent religion from first principles for at least a decade now.

Same. I've scrubbed my presence as much as possible from the majority of the internet (except HN for whatever reason) to prevent future models from being trained on my output, but the damage is already done. Part of me exists in pretty much all AI models now, without my consent. And those models are actively stealing my career and passions.

And the cruel irony is that they stole our work to train the AI that devalues our work going forward, and will cause many of us to lose our jobs.

I regret every line of open source code I ever wrote.

Is this "escape route" in the room with us now?

This was an incredible video, and I feel so many complex emotions after watching it. I wonder where everyone in that video is now.

Sounds like there's some controversy about this video though. There was some accusations that it was staged, and / or that this was NOT the first contact this tribe had.

The narration in the video says its tobacco

I've never once gotten that impression from him

There is legit grief associated with what AI is doing to our craft, and we're all trying to grapple with it in our own ways

Hopefully their advice includes: "take this advice, and then don't join the Jehovah's Witnesses. It's a high control abusive religion that will make your depression worse".

I often feel like this as well. It's hard to talk about it with people, because they panic and immediately start sending you literature about suicidality and giving you hotlines, etc, etc, as if I haven't already seen all of that before.

I'm not actively suicidal or anything (and I know all the signs of risk, and what to do about it), but there's an underlying feeling that life just hurts and has no meaning. Currently the thing that helps me stick around is the idea of "doing it for the plot". Life ends in short order either way, so I figure I'll just stick around and keep it interesting.

Yes! Claude was so pleasant to use at first. It was Anthropic's biggest advantage. And now it's like nails on a chalkboard.

> More work to be done (and we will!) but reading better prose makes me so much happier.

I assume this work will be done for Opus as well? Opus has seemingly gotten progressively worse at its prose and technical writing with each version. I've stopped using Claude entirely for now, because it manages to turn even the simplest technical explanation into the most obtuse and obfuscated word salad imaginable. People originally adopted Claude because it felt pleasant to use in comparison to ChatGPT, but I feel like that's really been lost (at least with the Opus line).

I feel dread when I see a wall of text generated by Opus. Every developer I've talked to feels similarly right now.

Or Auvelity, which is a newer one that mixes Bupropion and DXM.

I am a type 1 diabetic and this is not true, and an incredible oversimplification. And categorizing "very low carb" as "good" and "very high carb" as "bad" is bizarre and untrue. You can go into DKA on a low carb diet.

And low carb vs high carb for T1D is NOT a "good vs bad" equation to start with.

Agreed. I think PKD himself got trapped by words and trying to intellectualize his experiences. He experienced a lot of anguish because of that

What I love is how PKD and Horselover Fat phase in and out of each other throughout the book. Near the beginning, PKD makes it clear that Fat is "him" and that he's narrating the book to give himself clarity on the events. But then PKD the narrator slowly instatiates into the story as a completely separate character. There are scenes with Dick, Fat, and his friends all talking. And then that changes again later in the book.

I see almost 4 versions of PKD in VALIS: the narrator, Fat, the narrator instantiated into the story itself, and then the real life author (who is treating this book as semi fictional, semi autobiographical).

> who dives into the world of his comatose boss' brain

This isn't quite what happens. It's actually left extremely ambiguous what's happening and where they are after the explosion.

I have a copy of his Exegesis, and I can confirm, he reinterpreted the experience probably hundreds of times. He loved the gnostic interpretation (which is the main focus on VALIS). But also yeah, sometimes just straight up orthodox Christian ones. And he even fell back to interpreting it as just psychosis sometimes.

PKD also invokes Buddhism and Hinduism frequently in the Exegesis. Here's a quote I love from it:

> I have seen the infinities of Judaism, which is morality, of Christianity, which is love, of the Greeks, which is wisdom, and I have seen God’s power as pronoia and charis to rescue me by bending the world itself; but beauty is a perplexing infinity, raising more questions than it answers. It is a puzzle too intricate for me. It spans all else. As I sit across the game board from Krishna I say, “I have found in beauty that which I could not myself have made; thus I have found the benchmarks. I believe, for I have the evidence that I trust; it is sufficient.” There is an infinity of good, of love, of wisdom, of power, but each particular beautiful thing is infinitely beautiful, and there is an infinity of them, so beauty, alone, is an infinity of infinities: ∞2.

Sometimes he wrote his worst trash in between some of his best books. He was ultimately a pulp fiction author, and often just pushed stuff out for a paycheck.

PKD remained surprisingly lucid and self aware of the likelihood of suffering from some sort of mental illness. He went back and forth the rest of his life trying to figure out whether he was insane or whether he'd had "real" spiritual experiences. But as the characters in his books learn, there may not even be a clear distinction between the two.

From VALIS (which was largely based on his experiences):

> What he did not know then is that it is sometimes an appropriate response to reality to go insane.

> Let it be said that one of the first symptoms of psychosis is that the person feels perhaps he is becoming psychotic. It is another Chinese finger trap. You cannot think about it without becoming part of it. By thinking about madness, Horselover Fat slipped by degrees into madness.

> People suffering nervous breakdowns often do a lot of research, to find explanations for what they are undergoing. The research, of course, fails.

> What struck me was the oddity of a lunatic discounting his hallucinations in this sophisticated manner; Fat had intellectually dealt himself out of the game of madness while still enjoying its sights and sounds.

> It lacked depth and complexity in some ways that modern games have

I would argue it has way more depth and complexity than the majority of modern games. There were so many systems that interacted in interesting ways, and so much player freedom that the devs actually accounted for. Just endless stories of "I tried this bizarre thing that I thought nobody would think of, and the devs actually wrote unique dialogue for this situation".

It's also a bit of a mess and cobbled together, but in a beautiful way.

Once again proof that SOC2 is nothing but a marketing tactic, and busywork

> tl;dv names their microservices after pasta. A subdomain scan reveals cappellini, carbonara, fusilli, pasta, penne, puttanesca-v0, and ravioli, all under tldv.io. An entire Italian restaurant worth of Express servers.

Pretty appropriate, given a vulnerability of this severity. Literal microservice spaghetti.

(Also, please, let's all move back to boring names for services and servers. Nobody likes trying to decode what all these silly names mean.)

EDIT:

omg, the disclosure communication is infuriating.

> We're on it. It needs some time, but rest assured we're following through. For further communication, i'll recommend reaching out to our CTO

This should have been a P1 that was fixed same day, and they strung him along for months. Absolute amateurs.