Thinking loudly about networked beings. Commonist. Projektionsfläche. License: CC-BY
2693 stories
·
145 followers

Tech World Vibe Check

1 Comment

(This is going to be a quick one. I owe someone edits right now, and he reads this newsletter.)

The vibes surrounding Silicon Valley and the digital future are absolute poison right now. It’s so very weird. It’s so very bad.

Where to begin…

(1)The CEOs and venture capitalists want their Jensen Huang back. I don’t think this will work out for them. It would be ludicrous if it did. We live in ludicrous times. I dunno.

when one ceo is in distress, the others in the herd will adopt his distinctive plumage

— derek guy (@dieworkwear.bsky.social)
6:34 AM • Nov 19, 2026

One point I’d like to underline, boldface, and put in 48-point font: we should understand this as nothing less than a full-scale attempt at freeing the tech oligarchs from any conceivable checks on their power.

Huang was fired by his Board of Directors. For sleeping with a key business parter (who might be a spy for the Chinese government). And also for sleeping with two of his direct reports. That is… very much the sort of thing that gets a CEO fired.

It is also very much not the sort of thing that gets in the way of other tech oligarchs doing whatever they want. Elon Musk’s Board of Directors couldn’t fire him over this (lol). Mark Zuckerberg’s Board of Directors couldn’t fire him over this. Dario Amodei’s Board of Directors could fire Dario for offending the incipient Machine Gods, but they almost certainly couldn’t fire him over something so mundane and human as this! Sam Altman’s Board of Directors… oh come on, who are we even kidding?

The argument for bringing Huang back is, essentially: (1) he makes so much money for the company. Look how number go up. (2) All the other AI companies rely on NVIDIA acting like the First Bank of AI. Firing Huang might make other peoples’ number go down. And, also, (3) he’s too rich and too important for mere corporate governance rules to apply.

This was a problem when Sam Altman came back. It has been a problem with all the other tech barons setting up governance structures that render them virtually untouchable. And it is absolutely driving this little rally-round-the-leather-jacket effort today. If Nvidia’s board folds and brings Huang back, that effectively means Huang is untouchable. Too rich to be constrained, even by the most straightforward of corporate governance mechanisms.

(2) Anthropic went public today. It was the largest IPO in history, replacing SpaceX earlier this year as the largest IPO in history. Anthropic has had two profitable quarters, coinciding with the tokenmaxxing fad.

Investors aren’t ACTUALLY judging Anthropic’s share price on the basis of long-term expected revenues. (Ell—and I cannot stress this enough—Oh Ell.) That simply isn’t how markets work anymore, if indeed it ever was. (It’s a Keynesian Beauty Contest. Always has been.) Investors are instead betting that in the short term, a share of Anthropic stock will be worth more than it is worth today. And they certainly might be correct. Sophisticated investors love bubbles, so long as they can time their exit appropriately. Retail investors love bubbles too, believing that surely someone else will be the sucker of last resort.

One related point: Anthropic’s seven cofounders are now worth about $18 billion apiece. All the other early investors (who pretty much universally hail from Effective Altruism circles) made out like absolute bandits as well. Today’s IPO is going to mint an entire new generation of extremely-strange-billionaires. They are going to be happy to donate heavily to nonprofits that fit their worldview and political program. This is going to bend the progressive political advocacy sector in ways that will probably drive me, personally, insane.

(3) Jacob Coxon, the Anthropic whistleblower reveals himself to be such a little twerp.

Remember when Jacob Coxon quit his job at Anthropic and warned that the company might destroy the entire world? That was two months ago.

Anyway, yeah, Coxon was recently asked why he still has stock in OpenAI. His answer is worth viewing in full. He explains (a) he keeps the stock because “I like money bro,” (b) he might go back and work for one of these labs, who knows!, and (c ) he agrees that donating the stock would add moral clarity to his argument, but people could just stop asking him about it instead.

These are not the actions of a serious person. These are the actions of a 27-year-old, deeply embedded in a creepy millenarian subculture, who wanted to impress his friends and be famous for a bit.

We would be much better off if we learned not to take these people seriously.

Jacob Coxon, former researcher at OpenAI and Anthropic who quit and said that AI "could kill us all by the end of the decade" defended keeping his OpenAI stock this week by saying "I like money bro."

A clip of the full Stanford HAI talk was shared by Alex Kantrowitz on X.

— 404 Media (@404media.co)
6:36 PM • Nov 16, 2026

(4) Everything is breaking, and everything will keep on breaking. Microsoft. The airlines. The banks. The NSA. Just… all of it. This is some mix of (1) rampant AI systems finding and exploiting vulnerabilities, (2) vibe-coded AI junk getting rushed into production, then breaking upon contact with the real world, and (3) governments and companies shedding layer upon layer of competent employees, leading to the breakdown of critical systems.

My impression is that it’s mostly option 3. 2025 was the year of DOGE and tech companies fucking around. Now, in late 2026, we have entered a period of finding out. The nascent machine gods aren’t taking down critical infrastructure. The idiot billionaires and the government they helped install are causing all that damage themselves, by mistake. Even when the Dems retake one branch of government, it’ll still take years to sort out the mess before we can hope to begin repairs.

(5) Alex Karp says Palantir is totally progressive. Elon Musk says he was never MAGA.

Not to brag, but I saw this reverse-heel turn coming a mile away. Remember when Palantir posted that MAGA manifesto? That was less than seven months ago. I predicted at the time that, if and when the Democrats came to power, Palantir would respond by simply pretending the company had never posted it. This is one of the most dangerous and powerful defense contractors on earth, and they effectively operate quarter-by-quarter, telling the centers of power whatever it thinks the centers of power wants to hear. They are not capable of planning more than six months out.

I keep seeing people online ask whether the tech billionaires really thought MAGA would rein forever. And, not to be pedantic about the matter, but… no, that’s not what was going on there. The tech billionaires simply didn’t think that far ahead.

In early 2025, Elon Musk’s DOGE did irreparable damage to the federal government, and Musk cemented his status as a MAGA icon.

Now, in the least-convincing rebrand since Alex Karp last week, Musk claims he’s not MAGA at all. Read more about Elon’s politics here: www.wired.com/story/far-ri...

— WIRED (@wired.com)
1:15 AM • Nov 19, 2026

One important thing to understand about the tech barons is that, having conquered startupland, they fundamentally believe that everything in the world is vaguely startup-shaped. And startups don’t succeed through foresight or careful long-term planning. These are people who embrace extreme flexibility, “pivoting” away from products, business models, and belief systems whenever it seems convenient to do so.

The core conviction uniting the tech oligarchs isn’t any traditional political philosophy. Their core, shared belief is that they are all special main characters (founders!) who are singularly responsible for building the future, and that the rules shouldn’t apply to them. Jensen Huang, constrained by a board? Outrageous! Palantir, having tried to manufacture an authoritarian surveillance state, facing potential scrutiny from regulators? No no, you see that was the OLD Palantir! You can’t go holding Alex responsible for his past statements and actions!

It is important to ridicule them, lest they get away with it. They have such an abundance of resources that they’ll try to get away with it anyway.

It’s rotten. All of it. One of the central tensions running through this current historical moment is that “tech” is both the material technologies that we move through life with (good, sometimes, except all the breakdowns and enshittification) and also an assortment of the worst, most audacious capitalist shitheels the world has ever known (uniformly bad, and also inescapable).

The capitalist shitheels are getting shittier. Everything is falling apart. And their one demand is that everyone let them do whatever they want without consequences.

The sooner we make our way through this particular status quo, the better.



Read the whole story
tante
2 hours ago
reply
"Anthropic’s seven cofounders are now worth about $18 billion apiece. All the other early investors (who pretty much universally hail from Effective Altruism circles) made out like absolute bandits as well. Today’s IPO is going to mint an entire new generation of extremely-strange-billionaires. They are going to be happy to donate heavily to nonprofits that fit their worldview and political program. This is going to bend the progressive political advocacy sector in ways that will probably drive me, personally, insane."
Berlin/Germany
Share this story
Delete

On Moving Goalposts

1 Comment

For the longest time the responsible narrative around “AI” usage was: You can use AI as support but you really have to check all responses for correctness.

This was a very obvious attempt to address one of the most pressing issues with “AI”: The way it lowers the quality by introducing random errors. Surely nobody would have an issue with people using “AI” if they checked all the outputs, right?

Well of course people do have issues: Disenfranchisement of Labor, ecological impact of energy and water use as well as the creation of e-waste, the massive illegal accumulation of everyone’s work products and the negative impact on human cognition and flourishing should be more than enough to stop basically everyone from using those systems but ethics are not in fashion.

But if we stayed in the capitalist logic of the end/output/productivity justifying the means this does – at first glance – look sorta reasonable? But what always irked me was how this did not correspond to my experience at all: Reading, editing, validating someone else’s work takes a lot of time, especially when you need to be focused enough to catch those kinds of weird errors that LLMs introduce. I never understood where the time savings were supposed to come from. I did have a suspicion though.

Ketan Joshi has created an infographic outlining what I thought was going on:

"Joshi's Workslop Postulate"

It shows an image of a man in front of a screen saying "I use AI all the time but I always check the outputs"

The lower part of the image is split in two panes:
The first says "Either lying:
Ridden with heaps of catastrophic errors.

Definitely quicker than doing actual work I guess"
The other one says:
"or foolish: checked and corrected every single error manually.
Took longer to correct errors than just doing the work lol."

"Always check the outputs for errors" = "please recycle" for information pollution rather than plastic polution

But I recently see a shift in how “AI” proponents are talking about their usage. Because where the old (probably bogus) approach at least claimed to care about quality and keeping it somewhat up, we now don’t even get that promise anymore.

I keep seeing in my environment and hearing from other people more and more stories of “We don’t need to check things, if something’s broken, the AI will just fix it” or “we just use an AI to check the results”. Two stochastic wrongs maybe creating a right seemingly being the hopes and dreams and thoughts and prayers.

This is a massive shifting of the goalposts in “AI” defense. Which in general is a weird thing to do, because why defend something that unethical but it did try to claim not to make the world worse through creating shittier artifacts. But now we’re just full speed ahead into slop-world. What if Fast Fashion but for everything?

I think this shows us a handful of things. First it illustrates how “AI” (including those set up in “responsible” ways) cannot provide both: Productivity gains as well as consistent quality. It is a confession of sorts.

Second it does also illustrate how far into capitalist accelerationalism we’re leaning in general: If something is just faster, quickly satisfying some of our impulse that can be framed as a positive without being laughed of the stage or slapped in the face.

Third it shows that the “AI” crowd will never commit to anything and that all claims and promises from that camp are moot. They are running from society always claiming to provide something and never providing it. Because everything changes so much … innovation … old paradigms … bla bla bla.

I think it’s important to cast a light on this shift, on how all these systems were funded, developed and deployed under the promise of ensuring quality and responsibility and now that it has become obvious that that would never work, we’re slowly but surely narratively prepared to accept more and more slop in our lives.

But we must not.

Many of the criticisms against “AI” (exploitation, lack of regard for externalities, etc.) can also be made against most capitalist forms of production. “AI” isn’t special in that regard. But that weird asterisk that accompanies every “AI” system pushing all responsibility on the users who get put under more and more pressure to provide efficiency gains that “AI” cannot deliver must be put at the center of criticism within mainstream discourse – a discourse that sadly often also doesn’t care about ethics or environmental impact or people’s dignity.

Accepting slop means accepting objectively worse digital objects – produced at incredible costs to the environment, our societies, our cognitive and mental health. And that is where I think one of the first lines needs to be drawn.

California just sued OpenAI over them using stochastic systems to attack other organizations’ websites and platforms (it is very important to not adapt the framing of “their AI attacking” because that’s 100% not what is happening). This is another step of undermining the framing of “AI” being special, needing special rules and not having to obey to established norms and laws. Norms like for example you as a service provider having to reliably provide the service you promised.

If you apply that kind of common sense understanding of services to “AI” based services, none of them will survive. So let’s get to it.

Read the whole story
tante
2 days ago
reply
"This is a massive shifting of the goalposts in “AI” defense. [...] now we’re just full speed ahead into slop-world. What if Fast Fashion but for everything?"
Berlin/Germany
Share this story
Delete

The Myth of Neutral Code

1 Comment

Comments

Read the whole story
tante
8 days ago
reply
"Software is built by people, maintained by people, and hosted on infrastructure funded by people. Much like art, it cannot and does not exist in a vacuum."
Berlin/Germany
Share this story
Delete

Why You Should Care About The Political Project of Twisted

1 Comment

Twisted was originally the network component of a text-based role-playing game engine. In the late 90s when it was first developed, the idea of a video game being able to send you actual emails and instant messages1 seemed pretty neat, rather than the valence that it would have today (nightmare). We never ended up making an actual game, so we eventually removed the “game engine” piece, and just kept the protocol implementations of the email and the AOL Instant Messenger, and the DNS and eventually SSH and so on. The residual pieces of that project have not seen maintenance for a decade.

So it’s not really a way to make a text-based game any more.

Then, Twisted was the premier longest-lived and best-maintained event-driven I/O library for Python. If you wanted to make high-performance servers, it was the best and most obvious way to do it. Even if it still acquits itself relatively well in this regard, its claim to fame, and to indispensability, has been somewhat obviated by the inclusion of asyncio in the standard library. Even if it has some stuff to recommend it — better-than-average testing facilities, for example — you can get better-documented and better-integrated low-level asynchronous networking tools without even installing any packages.

So it’s not really the premier choice for an event-loop these days.

Twisted does still contain an SSH server and client, an email server and client, a DNS server and client, a web server and client (now including websockets). But you can cobble these together by installing a variety of packages from other sources.

So... it’s a bunch of protocols? And I guess it doesn’t use AI? But... why would anyone care about that? Why contribute to such a thing?

What is Twisted doing, and who is it doing it for? What is the political project of Twisted, nigh on 30 years into its existence?

Most of all, why should you care about it, or want to contribute?

If you’re asking what you could use Twisted for today, it’s an excellent choice if you’re implementing a new network service and you need more than one protocol. If you need to do HTTP, DNS, packet capture, and a custom SSH server, it would be very hard to arrange a collage of other Python packages to manage this. It’s still much better at deterministic testing than most of the standard library. It’s a bit easier to hook into a GUI.

But that’s not what Twisted is doing.

What I want Twisted to do for the world is still an unfinished project. It’s honestly kind of bad at doing what I want it to, but it could be great. The overarching goal of having and maintaining all this stuff is to, as much as possible, demystify networked software operation and let you build Internet facing systems at a human scale.

It’s all in a memory-safe language with good error handling, so you can avoid the scariest security pitfalls. It’s all integrated into a single paradigm with built-in tools for testing. It’s designed to scale down to be runnable on the average laptop, and not require large elaborate server farms or distributed clusters to run. Via the twist command line utility, it provides a Swiss army knife of simple tools that you can use out of the box.

In a world where all of our compute is in massive distant opaque datacenters, where a tiny cartel of gigantic tech companies control nearly the whole supply chain, one of the few things that you still can have agency over is your own server on the Internet.

It’s still possible to get a domain name for a few bucks a year, a free-tier or very low-cost VPS, then stick a glue record in and run your own DNS. From that point forward, you can run your own email, your own web server, your own terminal server. But doing this requires deep expertise, and that expertise requirement is massively inflated by the operational concerns that go with running twenty different pieces of software from twenty different projects, probably in twenty different languages, with completely inconsistent aesthetics.

The reason that Twisted is event-driven is because as much as it can be a pain to wrap your head around async/await (or worse yet, Deferred directly), it is much easier for a single person to understand the concurrency requirements of an async I/O multiplexed system than to understand what is going on with an application that requires multiple thread pools. It is possible to understand concurrent code that manages parallel tasks in that idiom. The reason that Twisted integrates so many different protocol implementations and attempts2 to unify them with a common authentication layer is to reduce the overhead required to manage a comprehensive suite of network services.

Even the not-using-AI thing is not an aside due to some accidental personal preference about tools — it’s a core part of building an engine that you can personally understand if you dive into it, that’s built with a consistent style and layered in a way that you can take apart if you need to. It’s not built with AI because we’re not building for AI, we’re building for people and the entire point is that people are supposed to be able to understand it.

We used to have a tag line, which is somewhat less emphasized these days because its meaning is not immediately clear without this background, but was once quite prominently featured on the home page: “Twisted: Building the Engine of Your Internet”. That’s what I want to bring to you: an engine for you to build your Internet, that is small enough that you can make it personal, that doesn’t require massive investment in management tooling just to get started.

Now, for various reasons, I no longer use twisted.names to host my own DNS (although I did for many years!), and we never wrote enough of a twisted.mail suite for it to be usable for email hosting on its own. But, if you’re reading this article, it was probably delivered to you directly via Twisted’s own twisted.web HTTP server, using twisted.protocols.tls to configure our own OpenSSL TLS termination. I have run with extremely high reliability for years and next to zero management overhead, in a single simple docker container. Which I then administer and upload files to via Twisted’s own twisted.conch SSH server.

If that seems like an appealing project and you’d like to volunteer to help out, drop me a line and maybe I can point you at some tasks to get started. Twisted falls far short of this aspiration despite 30 years of work, and we could use all the help we can get.


Acknowledgments

Thank you to my patrons who are supporting my writing on this blog. If you like what you’ve read here and you’d like to read more of it, or you’d like to support my various open-source endeavors, you can support my work as a sponsor!


  1. Yes, it was the 90s, we didn’t have “texts” yet, it was all “IMs”. ↩

  2. — with varying degrees of success — ↩

Read the whole story
tante
14 days ago
reply
"In a world where all of our compute is in massive distant opaque datacenters, where a tiny cartel of gigantic tech companies control nearly the whole supply chain, one of the few things that you still can have agency over is your own server on the Internet."
Berlin/Germany
Share this story
Delete

Remember Remember

1 Comment
Remember Remember

Oh no. It's November and I never finished the blog post I'd been trying to write about Halloween and how ghosts haunt everything and how haunting is good. How a song is a kind of spectre and how we are all summoners. About how art is an assemblage of hauntings that unravel the impossibility of authorship, and how to be an artist is not to make a claim on any particular work, but to be a curator and custodian of ghosts. To look closely at any artifact is to understand it is inscriptions all the way down. It would have been about how there is still a need for the artist despite all that, because somebody has to build a place into which we can summon our ghosts, a place for them to haunt us and each other.

In contrast, genAI, LLM-generated art is exorcised art. It is all surface. No friction, no demons, no cracks for the ghosts to inhabit. Does this mean that capital is an exorcist? If so then our task is to become summoners, re-enchanters, spectre-mancers. To reach beyond the world. Or, rather, to reach through the slick, digital froth congealing atop our world, this uncanny, skin-deep mimic of reality, grasp the dirt of the Real lurking in the shadows below, and rip it free of its illusory chains.

I heard mention recently of the 'intentional fallacy', which I hadn't come across before. I gather it refers to the mistake of thinking that when approaching a piece of art, deducing the artist's intentions is what is important. WRONG! What is important is what the art can do or, perhaps better, what you can do with the art. I enjoy the Jamesonian nature of this. The idea that we're all bringing all of our baggage (our albatrosses, our ghosts) to everything that we read or look at or listen to and it’s through and with this baggage that any and all meaning is created.

To me it suggests that all these pieces of art are a kind of communism, fragments of a collective conversation written in artifacts and scattered across history, making a mockery of temporality, talking back and forth as if believing that time only flows in one direction is a lack of imagination on our part.

But then what to make of intentionality from the perspective of trying to make music? I struggle to explain the things I make to myself in anything so cumbersome and explicit as language, but I still need to feel like there is some ineffable thing I am attempting to communicate. Otherwise why would I share it?

The answer, I suppose, is the same answer as usual: embrace the dialectic. Expressing intentionality through music is impossible; making music without intention is to make dead, worthless music. Both are true. Deal with it. So many of my conversations these days end up in confusions like this. I was no fun at parties even before a decade or two of dialectics and Marx - you should see me now!

WHATEVS. This is really only a problem if you think about it too hard. And you don't have to think about music if you don't want to. You can just make it. Or do whatever you want with it. Still. I fell into the realms of over-thinking ages ago and cannot climb out. Dwelling down here can really take its toll on the ability to easily access a sense of uncritical wonder.

Lately I find myself looking for a way to carve out some space to experience music as the force I know it in my bones to be, but without needing every song to sound like the howling of the end of the world, or to be in a room shaking with colossal noise to get there. Because apart from anything else, this tinnitus is not getting any quieter.

This, my bewildering quest for what I am reluctantly calling 're-enchantment' has been ongoing for a lot of this year. A search for a materialist-flavoured magic. I foolishly want to regain access to a power that has the ability to end me in an instant. For a song to crumble me to dust: a thing that I swear has happened so many times in my life that I cannot understand how I am anything more than vapour at this point.

Don't get me wrong though, I also do absolutely want to continue finding ways to make every song sound like the howling end of the world. Not because I need to, but because IT IS COOL.

But Halloween came and went and I never managed to pull all of this together.

Never mind. The Gothic Marxism of Halloween is good, but the celebrating-direct-action-ness of bonfire night is better. Happy almost-blowing-up-parliament-day to all those who celebrate.

Read the whole story
tante
14 days ago
reply
"About how art is an assemblage of hauntings that unravel the impossibility of authorship, and how to be an artist is not to make a claim on any particular work, but to be a curator and custodian of ghosts. […] In contrast, genAI, LLM-generated art is exorcised art. It is all surface. No friction, no demons, no cracks for the ghosts to inhabit. Does this mean that capital is an exorcist? If so then our task is to become summoners, re-enchanters, spectre-mancers."

Paul is really onto something here.
Berlin/Germany
Share this story
Delete

KI-Taskforce der Regierung: Die rosarote Vision einer deutschen KI-Nation

1 Comment
Die KI-Taskforce der Bundesregierung hat ihren ersten Bericht vorgelegt. Sie träumt von einer Art KI-Paradies, in dem alle von der Technik profitieren. Eine Analyse von Friedhelm Greis (KI, Politik)
Read the whole story
tante
16 days ago
reply
"Die Taskforce scheint das Ziel zu verfolgen, KI um fast jeden Preis in allen erdenklichen gesellschaftlichen und wirtschaftlichen Bereichen durchzudrücken. Das soll zu Wirtschaftswachstum, zukunftsfesten Arbeitsplätzen und einer modernen Verwaltung führen – einer Art KI-Paradies."
Berlin/Germany
Share this story
Delete
Next Page of Stories