Some thoughts on AI: A snapshot in a dynamic environment
As a software engineer, I am probably part of one of the first groups that really feels what AI can do. Especially with agentic AI, the progress during the last few months was extremely fast. I am basically using it all day long at work. On the one hand, this is impressive, but on the other, to be honest, also a bit frightening.
When I read the recently published Meta's manifesto on AI, it made me want to write down my own, current thoughts in this very dynamic environment.
I like the idea of AI making individuals more capable. But I also notice how easily I let it take over parts of my own thinking and reasoning.
AI is completely changing the way I work
I wrote about building my first application with agentic AI in September 2025, only one year ago. The workflow at that time is already outdated. I let the AI solve small implementation tasks; right now everyone talks about spec-driven development. Definitely, one year ago, I could not have imagined today's capabilities of coding agents.
I use AI because it is such a productivity booster and an easy way to brainstorm. But I also notice the other side. If a tool summarizes something, I sometimes read less carefully. If an agent creates a first draft, I start from there instead of struggling with the empty page. If every small problem has an instant helper (chat window), we might get worse at staying with hard things.
The problem is that this happens continuously in small steps. Each step feels normal after a short time. Then the next adoption happens. It is hard to notice how much behavior has changed unless we actively compare it across a longer time frame.
What do I still need to learn myself?
I am very happy that I don't need to do boring tasks manually. A lot of work should definitely be automated. But there is a difference between removing nasty friction and removing the small challenges that make us capable.
Education is the part where I am most unsure. I don't dare to make a strong prediction. Which skills are still relevant when AI is always available? How deeply do engineers need to understand the basics? Are there going to be new basics? If the old basics keep being important, how hard will it be to learn them when the shortcut is always there? I think we can't simply compare that to the invention of the calculator.
Too much comfort worries me
Another thought that stays with me comes from Mihaly Csikszentmihalyi's idea of flow. I recently read the book and I am currently a bit hyped on this topic.
What I found interesting is that people often experience flow more at work than during leisure. Work can provide structure, feedback, challenges, and the chance to use real capabilities. These are prerequisites for truly enjoyable activities (flow). AI should help with work, but I hope it does not steal all of that from us.
Leisure time instead is often just comfortable without being deeply enjoyable. Scrolling, watching videos, buying things. It can feel nice in the very moment (pleasure), but it does not always create the same satisfaction and enjoyment as doing something difficult and getting better at it.
Therefore, for me, one subtle AI risk is too much comfort. If AI gives people more time and that time goes into family, sport, learning, creativity, building, reading, and human interaction, I love that. If the extra time mostly becomes more screen time, more advertising and more Amazon shipments, I think we have already gone too far in this direction.
Job disruption will not stay abstract
I expect agentic AI to become broadly used inside companies soon. In startups as well as in big corporations. There will be AI tools and workflows not only for engineering but also for HR, finance, sales, and so on. Some will still require technical knowledge, but there will be many tools designed so that people can use them without being very technical.
Although it still feels weird to imagine agents as coworkers, some jobs may be completely replaced by AI. Who knows if, in 10 years, a large part of the economy will still involve work in front of a computer, but for the next couple of years I am sure that agents are joining us and doing some work with us.
I also hope that AI will create new tasks, jobs, and duties for humans that we cannot imagine right now. In the past, big technological revolutions often destroyed some work while creating new kinds of work. I think this will happen again. The impact on society will be defined by the timeframe within which this happens.
I like Meta's vision, but I question their incentives
What interests me about Meta's manifesto on AI is its focus on individuals and private life. I also want AI that helps individuals to become more capable, healthier, more creative, and less blocked by annoying work, but I struggle with Meta as the company behind that vision.
Meta has historically made a lot of money from attention, engagement, and personalized advertising. I actually use none of their products except WhatsApp. But from what I hear in my social circle, many people are not really happy about how much time they spend on social platforms. It is often described as a bad habit people want to get rid of. And I can understand that when I think about how much time I sometimes spend on YouTube.
When everyone has a personal agent, these systems may know our routines, goals, weaknesses, fears and relationships. I am sure most of Meta's revenue will also be made from advertising and not only from real subscriptions to great AI products.
The bigger picture is hard to predict
One thing I find difficult about AI is that nobody really knows where this goes. People talk about AI reaching or exceeding human intelligence, but no one can really know how this will affect the world we are living in.
The geopolitical part feels almost unavoidable. The US and China are the key players, and nobody wants to stop first if the other side continues. In game theory terms, this looks like a security dilemma or an arms race.
There are other big risks too: even more power for a few companies, bigger monopolies, and capital becoming way more relevant than labor and human capital.
This is all far away from what I can influence. I keep being optimistic.
My current view
I am completely hyped on AI and use it throughout the day in private and professional life. I am curious what comes next. But when I step back and think about it, my current view is mixed.
I want AI to help people work on things they care about. I don't want AI to lower our capabilities. I don't want AI to keep us engaged on advertising platforms or subtly try to sell us things we don't need.
For now, I want to continue using AI actively, but not blindly. I want to follow and use the latest tools and benefit from them. But I also want to notice when I outsource too much reasoning and choose comfort too often.
Learning how to use AI while keeping a fresh and capable mind sounds like a good challenge for the next couple of years.