Just a link drop of what’s caught my attention, from the political to the mundane “AI” company employee’s fare.

I’ve included this time my notes for each from my bookmark tracker; sometimes it’s the og:description, sometimes it’s an excerpt I’ve pulled, sometimes it’s my own notes, sometimes it’s nothing if I was just not feeling it!

67 links

• The Eternal Sloptember
  • I’m calling it now, the adoption of AI agents into software development will be one of the most costly mistakes in the field’s history. Agents cannot program, and it’s taking longer and longer to realize that they can’t. They are a highly sophisticated statistical model designed to mimic the distribution of programming. The output is broken, but in a way that’s getting harder and harder to detect. Which is exactly what you’d expect from an increasingly accurate statistical model.
• Clanker: A Word For The Machine
  • Why I like the word clanker and why machines are not people.
• Encyclical Letter of His Holiness Leo XIV: Magnifica Humanitas
  • From this follows a simple but compelling consequence: we cannot consider AI to be morally neutral. In reality, every technical tool embodies choices and priorities through what it measures, ignores and optimizes, and how it classifies people and situations. If a system is designed or used in a way that treats some lives as less worthy, or excludes them without the possibility of appeal, then it is not merely a tool “to be used well,” since it has already introduced criteria that contradict the inalienable dignity of the human person. For this reason, ethical discernment cannot be limited to asking whether we are using a system for good or bad purposes; it must also examine how that system is designed and what vision of the human person and society is embedded in the data and models that guide it.
• The Machine Stops
• The attack on competence
  • Understood in this framework, the LLM is simply the latest in a series of technologies that have promised the investor class that this time, finally, they can eliminate the need for competence and not have to consider or respect any value system other than their own.
• Cognitive Surrender
  • INCREDIBLE — ENTIRELY WRITTEN BY LLM. WOW. JUST WOW.
• Cave of Forgotten Dreams
• Quality in the Age of Slop
  • In the last third of the book, which is much better than the middle, Pirsig says that he wants to forget about all the metaphysics and investigate what Quality means for daily life. He says the most important thing that Quality can do is expand reason to include things that were previously inadmissible. This is important because it’s “the overwhelming presence of these irrational elements crying for assimilation that creates the present bad quality, the chaotic, disconnected spirit” of our modern age. The classical mode so dominates our thinking today that somehow I’ve decided my visceral, pre-intellectual distaste for AI is something I should discount rather than heed.
• To My Students
  • Lessons for a changed profession from a classic novel.
• LLMs and performative productivity
  • It’s worth asking whether LLMs are actually making us more productive at all—and if so, what we might be sacrificing in return.
• It doesn’t matter if it works
• our workplace LLM mass delusion
  • The in-house sessions showing off the tech are embarrassing; led by the wrong people, and should likely not even happen in the first place.
• Chatting with myself about AI
• Antidote
  • If you love building things, and the process of building is just as important to you as the result itself, it’s not unreasonable that you’re in a slump these…
• Human Routers of Machine Words
  • On people who use AI to think.
• If you let AI do your writing, I will come to your house and kill you
  • Did you think I wouldn’t be able to tell? I can tell.
• Extinction-level capitalism
• Is AI ruining our skills? Early results are in and they’re not good | Lobsters
• meaningless.ai
  • What this means is that the problems with AI will never be solved by making it “better”. In fact, the more successful it is at fooling humans into thinking it is meaningful, the more harmful it is. Just as counterfeit money will ultimately undermine an economy, counterfeit meaning will ultimately undermine meaning itself.
• No-One Escapes the Permanent Underclass
  • “We’re going to make this machine, and put it somewhere between God and the archangels, but also, it’s going to be as simple-mindedly obedient as a dog.” Does this sound like a good plan? Does this sound like the kind of thing that’s going to work out?
• Blink if you’re human
  • Or human-ish
• The Short Leash AI Coding Method For Beating Fable
• A global workspace in language models
  • Interpretability research on Claude’s internal thoughts.
• In defense of not understanding your codebase
• I love LLMs, I hate hype
  • I think from this blog you may misunderestimate how absolutely giddy I am about AI. I did hacking from 2007-2014, after that my whole career has been devoted to AI. I love the progress. I’m so excited for the new LLMs, self driving cars, video generation models, and coding agents. I set up a Linux box with opencode on my local GLM-5.2 last week and wow like just saying install tmux with the geohot configuration works; the Year of the Linux Desktop is finally here!
• Zig Creator Calls Spade a Spade, Anthropic Blows Smoke
  • Programming language drama ensues as tech bros oversell the end of software engineering.
• Know thine enemy: A critical engagement with AI-assisted software development
• Linus Torvalds on LLM usage in kernel development | Lobsters
  • The tools work, yes, but LLMs and derived tools are a comprehensive manifestation of the problems of the industry, many of which Linux was originally created to sidestep. It’s disappointing to see that he is now so easily swayed by the appearance of convenience, and his willingness to ignore the problems that lead others to not want to use these tools in the first place because of that convenience.
• Faulty Towers, vibe sickness, and the vibe bobsled
• Protecting our FLOSS commons from LLMs
  • (You bet I voted for this outcome!)
• Why Linus is Right and AI is Wrong
  • Here’s what they miss about AI criticism: every single thing
• Why Software Factories Fail
• A Voice from Nowhere
  • The remnants of human understanding in generated content.
• Beware the Permanent Periphery
  • The result of this protectionist loop might not look so bad at first. If you expect the next few years to be messy in many ways, you might think the middle powers would almost feel vindicated in the short term.
• Don’t Take the Black Pill
• AI: Considerations for people who make decisions
• Mathematics Without Mathematicians
  • We must know. We shall not know.
• As a Paradoxical Tool, The LLM Will Never Be Reliable
  • LLMs are failing to disappear into our most important practices
• help wanted
• rust-lang/rust is adopting an LLM policy
• Profession by Isaac Asimov (1957)
• The Contracting Circle
  • Why don’t we believe LLMs are people?
• Error by AI scribe during medical appointment leaves patient devastated
• I’m done using AI
• Every Fucking Website — Slop Edition
  • Every AI-generated landing page, all at once, labelling itself as it goes.
• What To Expect When You’re Expecting (A World of Multi-Agent Systems)
  • How do we model the future of work? It takes an AI village.
• Our Servants Will Do That For Us
  • Why the post-scarcity utopia is unlikely.
• Don’t paste the AI, please.
  • If someone asks you something, send your answer. Not the AI’s.
• AI Coding will Prevent Expertise
  • The need for ongoing friction in long-term skill formation.
• Changes to SourceHut’s terms of service regarding LLMs
• Creepy crawlies
• Stop Thinking of LLMs as Next-Token Predictors
• Why we should anthropomorphize AI agents
• The No-AI list
• The Other Road
  • The future I dreamt of is one where we have the choice to be more: to see more, live more, know more, do more. To build marvelous things. A world whose art and science is to ours as the Old Masters are to the coarse coinage of the dark ages. Not the obsessive, uncompromising, abstract monoculture of LessWrong utilitarianism (tiling the universe with synthetic soy-fed shrimp hedonium) but a plural future, human at every scale, from atoms to galaxies, where baseline humans can till the soil and swim in oceans under the light of stars veiled with Matrioshka brains; with thinking glass, ensouled glass. And humanity radiating into endless forms most beautiful—to storm Heaven, and abolish Death—and a sphere of living matter, centered on the Earth, expanding through the universe and turning cratered rock and ice into oceans and gardens, dead matter into living matter, breathing life into the dead world—to make the whole cosmos into a garden—“and rivers in the desert”.
• “Do You Still Read the Code?”
  • Avoiding interaction with the code is not progress.
• Maintaining the love for coding in the time of AI
• On Being Replaced by AI
• Peer review in the LLM (mania) age
• tokens too cheap to meter
• Plan mode is dead
• What Would A Serious AI Product Look Like?
• Beware: Real risks of using AI as a software dev team
  • An extensive list chronicling the risks associated with software development teams leaning hard into AI-generated code and related tooling.
• Is sandboxing sufficient to contain rogue agents?
• Dark Forestry
  • Seasons are changing for technologists, and the industry we work in isn’t the same as it was a decade ago. We have new realities and new responsibilities to accept.
• Cyberpunk is now
  • Finally!
• Programming Isn’t Special

¯\_(ツ)_/¯

If:

• you have your search domains set to x, y;
• you have connectivity on both of IPv[46];
• a.x only has an A record; and,
• a.y has both A and AAAA,

then ssh a — 100% reliably for me, YMMV — will try connecting to a.y over v6 first!

This has had me confused for days, having recently reëstablished v6 in the home.

(Concretely: search domains are meshnet, public; vps.meshnet only has an A record, vps.public has A + AAAA. The VPS only listens to SSH on the meshnet. Up until now, ssh vps always went via the mesh, but since getting v6 up at home, it’s always trying the public address via v6 first and so hanging. It does actually time out and then try the meshnet address, so I guess it’s resolving all the candidates itself.. which makes sense, SSH takes care around address/IP mappings wrt. known hosts etc. The easiest solution for me has been to just add the v6 mesh address to DNS!)

Second-opinion rheumatologist could only suggest retrying amitriptyline. Sure, what the hell, let’s try it again. No benefit, not even a passing improvement to sleep. Arrhythmia much more pronounced and RHR clearly elevated. Discontinued.

Now (finally) with a long COVID clinic. “Officially” “diagnosed” with LC; POTS and MCAS both described as likely. I feel the initial appointment was a bit of a form letter and that they’d say the same to anyone who bothered to get this far, but still; started low dose naltrexone, metformin, cetirizine, famotidine, all at once, ramping up the LDN and famotidine over a couple weeks.

Too tired to hope, but appreciating the new experiences nonetheless. Where there is change there is always a chance.

Anteriormente: One month at GitLab.

Selamat pagi! Cómo vuela el tiempo; la semana que viene será un año desde comencé mi trabajo en GitLab.

Mucho ha cambiado, y mucho no. Todas semanas todavía trabajo en seguridad, hago correcciones pequeñas, y mejoro todos tipos de cosas. Al mismo tiempo, hago muchos, muchos repasos, típicamente por más de la mitad de cada día¹, y me he convertido en mantenedore del backend de GitLab — ¡una de las experiencias más hermosas de mi carrera! —; diseño nuevas funcionalidades según mis propias necesidades de la plataforma (y mis caprichos del día); he comenzado a hacer turnos de guardia; he establecido varias relaciones en la compañía; me ha dado cuenta de que estoy trabajando para conseguir un ascenso — es decir, ¡que me importa lo suficiente como para desearlo! —; y, desafortunadamente pero insoslayablemente, he estado usando los LLM también². A pesar de esto, sigo encontrando formas significativas de disfrutar del trabajo.

¡Por el próximo año!

1. gracias a ya-sabes-qué ↩

2. Hasta el día de hoy están prohibidos en mis proyectos comunitarios; ganarse la vida es otra cosa, particularmente para las personas discapacitadas³. ↩

3. Todavía me niego a dejar uno escriba por mí o hable por mí, en cualquier lugar incluso del trabajo. Predigo firmemente que esto nunca cambiará. ↩

Esta semana nos mudamos de casa. Todo se ha vuelto silencioso y bonito. Estamos rodeadas de pájaros y árboles y el hermoso gran cielo. Tenemos espacio y seguridad. Fue mucho trabajo y aún queda mucho por hacer, pero por fin, por la primera vez en no sé cuánto tiempo, yo siento paz.