Competence as a virtue

Since LLMs became widely available, I’ve watched cognitive competence atrophy across domains—programming, writing, art. The pattern is clearest among less experienced practitioners, but no one is immune. We’re delegating our thinking to tools that deliver value faster, and in the process, we’re sidelining the very competence that creates genuine expertise.

I notice this degradation in myself. When I encounter a task that requires deep thinking, my brain feels sluggish, as though the neural pathways for sustained concentration have weakened from being stagnant. This isn’t surprising. Neural pathways that go unused atrophy. The more we outsource our thinking to LLMs, the less practiced we become at wrestling with difficult problems and eventually have difficulties reasoning about the simplest of tasks. The solution is what you’d expect: we need to regularly engage in cognitively demanding work that forces us to think deeply.

I’m not here to doom-monger about AI. As a programmer, I use LLMs a lot; to write boilerplate faster, clarify syntax, review my code changes. They’re genuinely useful tools that accelerate my work. But there’s a crucial distinction between using AI to move faster and using AI to avoid thinking entirely. The workflows differ by domain: what works for programming might not translate to writing or art, where the thinking process itself creates the value, the thinking in these domains is what produces the artistic elegance and profound messages that are carried through space and time.

One such example would be Raskolnikov, the protagonist of Dostoevsky’s “crime and punishment”, Raskolnikov is a character that is metaphysical. What he stands for surpasses space and time, you and me can embody what it means to live life like Raskolnikov, being tempted by Luciferian intellect and rationalized devious envy. This ability to write such stories, can only be attained through deliberate practice and deep thinking.

AI is a force multiplier, which means it amplifies what you already have. Imagine Dostoevsky with access to modern LLMs; he could have written faster, researched more efficiently, perhaps produced even more than he did. But give me the same tools, and I still won’t approach his level. The difference is competence. Dostoevsky spent decades developing his craft, learning to create characters that endure across generations and stories that explore the depths of human nature. AI can’t replace that foundation, it can only multiply what’s already there.

David Heinemeier Hansson, the creator of Ruby on Rails, recently described “competence as a virtue” in an interview. The phrase struck me because it names something I’ve been feeling: every time I use AI without understanding the details, I feel robbed, not just of cognitive competence, but of the depth I want in subjects I care about. Competence isn’t only about delivering value to the world. It’s about being at ease with yourself, knowing you’ve genuinely mastered something rather than assembled it from AI-generated parts that in turn give a vague feeling of productiveness and pseudo-competence.

What would it mean to pursue competence deliberately? To practice generating original ideas rather than prompting an AI? Whether you’re writing software, novels, creating photographs, or paintings, any domain that requires genuine thinking, the principle is the same: build real capability, not just facility with AI tools. And honestly it is incredibly hard, but this is what happened for generations and millennia, before you and I were born. Humans coming together and thinking deeply about a problem. This is what gave us the beautiful things that we enjoy today, the process with AI might just be slightly different and the iteration process a bit faster. But the fundamentals still remain the same.

Use LLMs, they’re powerful and they’re here to stay. But don’t delegate all your thinking to them. Treat them as companions that accelerate your work, not replacements for your own cognitive effort. The goal isn’t to reject these tools but to use them wisely: let AI handle the mechanical parts while you focus on the thinking that builds genuine competence. Because cognitive atrophy isn’t inevitable. It’s a choice we make every time we reach for an AI to think for us instead of with us. Build your fundamentals. Practice thinking deeply. The competence you develop will be yours in a way that AI-generated output never can be. Might even be useful to sometimes go for stretches of time where you do not use AI, this might also help you realize your true “competence” in a topic or domain.

In conclusion, strive for competence and deep understanding of the world that we live in. This will help you escape the cognitive corrosion that comes with over-reliance on AI.