Mastery in the Post-AI World

Sep 2026

Since I've started university, I've always felt that formal education would never be enough to become really good at my craft. I'm not even sure about what my craft is yet. I just know that I'm fascinated by technology and I've lost myself many times in trying to understand it. The more I explored, the more I understood how much there was to know and how little I could master. Eventually, I accepted the fact that I will never be the living encyclopedia I would like to be. At least not while I'm striving to become one.

A controversial aspect is that I have never taken much pleasure in becoming extremely good at one thing. Part of it is due to a desire to appear humble, but after a deeper introspection I believe that this is also because of a clear awareness of my falliblity. As a human being I cannot be consistently the best as a specialized machine is in its domain. I get tired, distracted, I misunderstand things.

The better you become at something, the more people start to address you as the one that knows everything about X: creating expectations of infallibility. But time is unforgiving, and the more you expose yourself, the faster you move from possibility to certainty that you will get something wrong, disrupting your reputation.

That's why every time I feel like I'm very good at something, I hit the diminishing return threshold of satisfaction, and further learning starts to feel like an opportunity cost. Especially now, with LLMs empowering anyone to do virtually anything, why should I specialize?

I still believe that mastery in a particular field is achievable only through countless hours of effort poured in, but now the persistent thought that I might be learning something else is holding me back from going all-in into a niche field. This is probably the best time in history to learn something completely new; even if you do so in an unstructured way and gain superficial knowledge, you might be able to produce insights and see opportunities that traditional deep specialization might miss. Breadth is more valuable than ever: LLMs will handle the technical implementation.

More formally, John Wentworth calls it exploiting dimensionality: you don't need to be the best in the world at any single skill if your combination of sub-optimal skills places you on a useful Pareto frontier. The more dimensions you can operate across, the more possible combinations there are and AI lowers dramatically the cost of exploring them ^[1].

On the other hand, you'll probably have shallow ideas to implement if you don't have experience in that field. I'm not talking about technical expertise: I mean a lot of time spent ruminating about a particular idea, a social matter, a service you use frequently... Only that allows you to find out what is worth contributing to.

The type of knowlege one needs to develop a complex epertise, like refined aesthetic sense or a mature judgement, is not simply a cognitive skill. It cannot be condensed into a formula: you could try to give a structure to your learning process, change your note-taking habits, adopt a different study technique, but this will only have marginal benefits. Achieving mastery is not a cumulative process. It requires a serious commitment of emotions and will, with an iterative trial and error process.

I see lots of people using AI trying to get answers immediately, using them as planning gurus to organize their lives. I'm not saying this is good or bad inherently, I just find it sad because you are missing out on the joy of disinterested passion. It is true that the cultural context changed and we have to adapt, but I cannot think of major inventions that were made with a linear approach. You start with a broad idea, you work it out and you branch off different paths, eventually ending up with something completely different from your starting point. Just as I did when I started writing this post.

The freedom we have now to start learning or doing something new is unprecedented in history. The sunken costs of failing to achieve your current goal are greatly amortized by the fact that you can start fresh immediately, catch up with the aid of a super smart, always available, artificial intelligent assistant and forget about your failures.

I'm currently reading Bauman and he describes freedom as

the measure of how much you can act according to your will obtaining results that are in line with your intentions^[2].

I find this definition particularly fascinating and incredibly relevant today, in the context of what we are discussing. We can now do virtually anything within the physical limits to which AI is restrained, but how many times are we tempted to accept the half bad answer of an LLM just because we are too lazy to give it enough context? We should always articulate our thoughts: for us and for the AI to effectively align with our intentions.

The real risk is not that AI will make knowledge or specialization worthless. It's that we might lose the patience to develop intentions of our own, because we are too surrounded by effortless answers.