Skip to contentšŸŒž
Back to notes

Thoughts on AI

Planted on: January 2, 2026

Recent growth on: September 15, 2026

Just like we had the internet bubble in the zeros, we now are deep in the AI bubble. I'm maybe even more so in this bubble because I'm working in tech and coding is/was a large part of my every workday. Honestly, I'm feeling quite conflicted about this entire topic.

How AI is affecting my working life

My work has gotten very different in the past two years. Some parts I like, others I don't. Coding is one of the fields where the use of LLM's in your work makes the most sense: output can be checked, the work is mathematical in a sense, and people are open to working with innovative technology.

It leads to me now In September 2026 always having three terminals open where a Claude chat is spinning (cool tech bro language for: I am talking with a chatbot in a nerdy interface). In these terminals I'm talking to Claude about my projects and tasks, and I'm more so guiding "it" to the right answers without me having to write any code myself. On the one hand, this makes it way faster to complete tasks and therefore do more work at once. On the other hand, it makes it by default harder to have a high form of quality control since you're likely to just trust the thing. And it also doesn't always make the work more fun: yes, you might be able to finish the boring repetitive stuff faster and therefore have more time to think about the cool stuff. But it also results in context switching often, which leads to mental fatigue, and in less satisfaction at the end of completing tasks.

So I'm not quite sure whether I still like my work if I still have to talk to Claude for the next 10 years. But I'm maybe even more worried about how LLM's are affecting other parts of our everyday life. Let's first rewind a bit and talk about what AI even is.

What is Generative or Agentic AI?

What the hell even is AI? Some people use quotation marks around the term "AI" since it is so out of context these days. AI is a pretty broad category, which has been around for decades, and basically just means any computer program trying to look like intelligence. I feel like the term AI is intentionally vague at this point. The way it is used currently is a marketing term above all. Even household appliances all have AI now, and are marketed as such.

Other forms of AI, such as predictive models (think of SARIMAX models: statistic models that use time-ordered data to forecast future events by taking into account seasonal cycles and external factors), have already been widespread for a long time: we use it for electricity demand predictions, traffic planning, finance and budgeting, etc.

But what everyone currently is talking about is another subset of AI: namely, Generative AI. Generative AI is based on large language models that are trained on so much text data to recognize patterns and produce human-like responses. 99% of the people who nowadays use the term AI actually mean generative AI or large language models - so I will as well in these notes. Agentic AI is another layer on top of that which can also execute tasks. So it has access to external systems and can therefore do things.

There are a lot of feelings and fears around the subject at the moment: there's undeniably a big hype by both tech people and everyday technology users because of either the rapid innovations and promises as the way it makes everyday tasks easier by eliminating manual and labor-intensive tasks. There is also fear of AI taking away jobs, taking away creativity and/or critical thinking. Of course, there are also valid concerns about how it impacts environment, politics and our culture as a whole.

A philosophical view on AI

Thereā€™s this great essayā¤“ by Hannah Baer in the latest issue of Artforum, where she unpacks why we are so afraid of AI becoming more intelligent than humans. If you look at this historically, humans have positioned themselves as the most intelligent species, even though thatā€™s not true. As the ā€œmost intelligent species,ā€ weā€™ve used this power as an excuse to dominate other people, religions, natureā€¦things that are seemingly ā€œless intelligent.ā€ If we continue this thought process, it makes sense that we might fear a more intelligent AI, as we could only imagine that it would try to dominate and extract resources from us. Itā€™s not about the tool, itā€™s about how the tool is used; itā€™s not about AI itself, but rather how AI will be used for big business, widening the class divide, etc.
(Mindy Seu in Passerby Magazineā¤“)

Diving into Baer's essay, she mentions how we are very quick to think about all the negative consequence that the popularity of AI has right now, while AI also raises philosophical questions about what it means to be able to draw, write and think. People in the creative industry have always struggled with the question of authorship. If AI is our creation, do we also control it?

Isn't it the case that because we as humans, especially in Western society, always want to improve ourselves, that we feed data into LLM's that match these narratives. Isn't the output of what we dislike from LLM's exactly what we are afraid of ourselves because of our feelings and values? The fact that quite some people are having relationship with chatbots right now, isn't that a mirror of the loneliness epidemic?

Technology is political. People like to think it's not about the tool, but how you use it. But the tool is created by people with very specific biases, intentions and expectations for the tool. And in many ways, once that's out in the world, people use the tool as designed.

I would like to unpack more of AI by using this note as a MOC (map of content) that links to facets of this topic.

Let's start with the positives

  • Eliminating repetitive tasks. Tasks that you can easily check the output of, and that no one really likes to do. Such as automating the spreadsheet that logs book orders, for grammar checks on languages I don't really speak, for grabbing information from multiple sources and summarizing these in one file.
  • The threshold for building software is lower for people with a lack of technical skills. If you have a cool idea, you no longer need a technical cofounder to build an MVP. It's way easier to build a website or an app. Think of people building a live score tracker for board game nights, a digital cookbook for family recipes, or a personal styling assistent. Some people call this the Age of Authorshipā¤“, because everyone is able to build stuff now. There are workshopsā¤“ now about how everyone can build real products from scratch with the help of Claude Code.

The negative impact

At work or in school:

  • AI (possibly) elevates the importance of human work behind genuine impact.
  • It also increasingly outsources a class of activities that give us satisfaction, accomplishment, and purpose. Michael Sacasas wrote about this in Do Not Resign From Lifeā¤“ where he emphasizes the importance of freeing ourselves from this existential angst to new technologies: our human uniqueness does not depend on a set of capabilities that we have as humans.
  • Burnout happens much faster in an AI-first business. Psychologically, people are so used to things going fast all the time, that they now have speed-based expectations that aren't healthy. There's less room for thinking about things deeply and making good decisions. When speed becomes normalized, we default towards fast / automatic / shallow System 1 thinking. Intentional, deep Systems 2 thinking, becomes a bottleneck.
  • Generative AI places a big emphasis on the end result and eliminates friction. For example: vibe coding is addictive because it gives you feedback and rewards like a video game. It removes the social cost of iteration. But it also for example leads to students cheat coding their way through difficult school topics.
  • The impact of AI on the education system in the sense that it removes friction for students, and amplifies the current system of performance pressure on both teachers and students.

Human intelligence:

  • Anti-intellectualism: As an outsider or a student, you might not notice that a large part of text is written by AI and is therefore a cheap imitation. "It sounds good", you might say. However, the teacher might disagree.
    • Less critical thinking
  • The impact of AI on language

In society

  • People wrestling with loneliness and isolation are turning to AI-powered apps. Same goes for (mental) health issues. The risk could be high.
  • Data sovereignty. Privacy. The fact that a few big companies can gain LOADS of personal data from LOADS of people.
  • Bias in data that LLM's use and (thus) The Gender Gap in AI - and what this means for women and other minorities
  • Companies are in a race with each other to make as much profit as possible. Within a capitalist model, you are basically doing things for profit with as little labor or low labor as possible. Therefore, this idea that AI will eradicate labour and redistribute leisure for the rest of us can never happen in a capitalistic model. It will just create new forms of labour that will be extraordinarily cheaper. We're replacing a lot of skilled labour with its likeness.
  • LLM's get good by using more data. This means more information being scraped from all over the place, and preferably human-made information to replicate in conversation to sound as human as possible. Unfortunately, this also means large-scale copyright breaking and stealing of other people's content. Let's say you're planning a trip with ChatGPT: well then AI has probably scraped data from a large number of travel bloggers without consent, credit, or compensation.

Open questions

  • How harmful is Generative AI on the environment?
  • How much impact will AI have on the labor force?
    • This is a double-sided sword. We need to do things more efficiently, especially in high-income countries because of the aging population. However, there are some things that might be argued are better off if humans do these things. Such as a psychologist being present in a conversation, offering you non-verbal empathy, and trying to figure out how what you are saying and what you are displaying say about the things you might need.
    • When AI prophecy failsā¤“
  • What will the future be of these big AI companies? AKA: The race for a profitable revenue model for AI companies
    • Digital colonialism (look for paper that belongs to this!) (tiktok video of Syed Faizan Raza): Generative AI is a loud solution searching for a problem, in many cases. (The toaster comparison! Load factor problems. Appliances needed to be invented to keep the grid running 24/7. Induced demand or better said: supply side desperation). AI companies have built massive data centers, and they need you to keep the meter running, because they are not generating profits yet. The fact that Whatsapp rolls out AI-generated summaries for private messages feels unnecessary and intrusive, because it is. Generative AI is being inserted into our daily lives, such that our habits will mach their financial projections. Software is no longer a tool, waiting in a shed until you need it, but you can now compare software to territory: AI agents are actively moving into our house. It is a sovereignty issue.

Reference

Back to notes