Unavoidable essay about AI
LLMs are a form of life, same as mathematics and the computers that make them. We interact with them, behave with them, and exist with them. An industrial economy is a form of life. That's all I will say on the topic of singularity/consciousness/superintelligence, as I believe it's senseless speculation regarding matters beyond anyone's control or epistemic capability.
The internet never needed a singularity or superintelligence to have a profound impact on our culture and lead to the near-extinction of many ways of life.
Personal
Early stories
My first encounter with LLMs was, like many people's, playing around with GPT-3 and being impressed with its sophistication. Of course, it wasn't yet a major cultural force, as many new technologies come along and we inevitably forget about them or treat them as unexciting parts of the background (such as 3D TV, VR, NFTs, OOP, the list goes on).
My encounter was fun and brief, and basically until late 2025, AI was this neat tool that helped me occasionally resolve problems that 'old internet' tools were insufficient for.
To me, it was a technology present in people's lives but without the capability to modify culture in significant ways (I would say the internet, spreadsheets, databases, synthesizers, transistors, and general digital media tooling are what I consider transformative and serious). AI, then, was far too crude to produce reliable results, and needing to use those aforementioned 'old internet' tools as primary references made it mainly a nice convenience rather than the powerful force it is now.
The watershed moment
But given that I am someone who does not avoid cultural osmosis, I learned of AI agents fairly late, probably in mid-2025, and still wasn't considering them seriously. Early adopters, though, are very efficient at transferring money to each other through their projects and discussions, while making much of society uncomfortable about the direction we're going. Regarding their work, I think it's best to quote one of the greatest English writers, Thomas De Quincey, in his wonderful Suspiria de Profundis:
He whose talk is of oxen will probably dream of oxen; and the condition of human life, which yokes so vast a majority to a daily experience incompatible with much elevation of thought, oftentimes neutralises the tone of grandeur in the reproductive faculty of dreaming, even for those whose minds are populous with solemn imagery. Habitually to dream magnificently, a man must have a constitutional determination to reverie.
Around that fascinating moment of Opus 4.5's release, I finally, and in my untimeliness, decided to give this (meaning agentic coding) a try. A basic implementation of something I had wanted for a long time: a cmus add-on to support Navidrome. And the great surprise: it worked. An advanced C project (with all its (n)curses), done in one evening. This was not a toy anymore; this changed things. Of course, the solutions it provided were inelegant, but guess what, I was never a good programmer, and frankly, I lack the patience to get good at programming. A shame.
In my private life, I often find the current technology limited by its own post-training and the difficulty of customising models. Maybe there's some future potential. We'll see...
In my professional life, though, these issues don't arise, as the tasks are fairly mainstream: creating dashboards that provide actionable business insights and developing quality tooling. Here, moving from chaotic and complex Bash scripts to elegant Go code is a complete transformation. I was aware, of course, that Go would solve a lot of issues in my field, but developer/admin time is simply very expensive. I'm frankly not excited enough by compiled languages to work and think about computers 16 hours a day...
Societal
Il n'y a pas de hors-texte
The development of computing technologies is deeply interconnected with the history of neoliberalism. We do not know what computing will look like outside of this world. But as people tie the technology itself to the power that holds it, they can be completely against it, even if the technology itself upsets the power that creates and holds it. I thought people were well read in Marx and understood that capitalism creates contradictions for itself, but it seems idealism is way too addictive.
On one hand, yes, these technologies are absolutely the tools of this failing machine of an economic system. They show up in absurd valuations, investments, bubbles, and other excitements. They are promoted in companies because companies need to show that they participate in the new, exciting form of capital development. This often makes many people's lives suck (to keep this lightly), and any tool given will still dance this dance.
On the other hand, what we are seeing with decomps (Deus Ex (2000) in my browser...), Adobe software clones, and other projects like these is the start of a revolution in computing. And this is with models that are adversarial to users and fairly limited. I believe the future of AI lies in local, much more customised, purpose-built, cheaper models and greater accessibility. And hence, I think the current titans will slowly devolve into the RCAs and AOLs of yesteryear.
Dreaming the same dream
One of the biggest accusations against AI is the environmental one. And it is correct, it is absolutely correct. It is correct like Foucault is correct, and it is weak because of its correctness.
Similarly, we cannot disregard Jevons paradox, but it's a long discussion. We have been very effective at reducing the cost of green energy, with the cost of electricity from utility-scale solar falling by 90% between 2010 and 2024. On the other hand, I think the effects of environmental changes already underway will make our lives more difficult, and some things will become more complex and require different models of industrial economies and societies.
AI plays a role here: it slowly enables new technologies, good and bad, needed and useless. It is also well loved in societies occupying different positions in our global economy. It democratises knowledge and learning, even if this happens through some gigantic data centre built by a corporation. The internet and its babies have always been strange lifeforms...
This environmental relationship to industrial growth is a form of management, and when management and compromises are needed, it means fundamental problems still exist. Maybe there will be a technology of Haber-Bosch-process importance that ends this entire discussion; if not, it will be a mixed process leading to the end of computing as we know it. We do not know what computing looks like in localised, syndicalised guilds; we do not know what computing looks like if scavenging is the main economic activity.
But as of now, we have to find ways of participation and non-participation, and all are expressions of this world.
To work is to forget
To start with, the industrial era hates inefficiencies, and if there's a cheaper way to do something, it will destroy some ways of life. It's a tragedy, and it has destroyed many lives. Of course, people change and adapt, and the value of a craft remains, but still, demand will falter.
As we speak of cognitive debt in this discourse, the relationship between knowledge and production is in a constant state of change. Once, it was necessary to know silver chemistry to take photographs. Today, to take great photographs, only aesthetic and technical taste are necessary, but some elements are forgotten.
To learn things is to know that we are not learning other things, and forgetting other things altogether.
Personal Perfection
To speak of people who pride themselves on computing without AI, as if this is some kind of virtue (besides the obvious virtue of craft): computing since the IBM PC has been deeply, deeply, deeply interconnected with (post)industrial inequalities. In many ways, there is little difference between a PowerMac G5 (which anyone in these microcomputing circles sees as a purely human, enjoyable retro project) and an H200 (besides, of course, technicalities). Both are peaks of their own logic and require economies of gigantic scale, involving a lot of dubious, highly complex, and intense processes.
To note, I personally believe our greatest work should have zero monetary value, as then only pure motivation remains. I think the Edokko way of life, in which you spend all you earn, was one of the highest and most meaningful approaches to living we have achieved. To paraphrase Bataille, the aim of life is to waste the energy of the sun.
And as for the accusation of AI slop against every project that uses AI, I say the only thing that defines whether something is slop or not is intention. If someone makes something purely for financial gain, it is slop, AI-assisted or not. If someone makes something with care, for personal use, and without much aim of gain, it cannot be slop (even if it makes you see things you don't like...). It remains human, as AI simply enables that human to achieve feats that were so far limited to the most time-rich people on Earth: 20-something in Omsk, Siberia. And it is our duty to become 20-somethings in Omsk, Siberia, and this hopelessness and instability are the greatest source of creative force and power there can be.
When dreams die, one is forced to confront realities and form new things.