Engineering talent and education

10,000 hours of coding have brought many software engineers to mastery. With AI came an abrupt end to the high influx of junior people, and the omnipresent questions “will we ever need junior people again?” and “how does a junior become a senior?”

I strongly believe we need a healthy number of juniors; it’s just that the way they reach mastery will be different. The more we trust LLMs, the less we correct their huge output, and the more we focus on architecture, on good practices, on high-level steering.

And mastery will still be reached as in any other field: through years of practice, for example taking over smaller live subsystems and owning the result — maintaining the system, updating it, failing of course along the way — and keeping at it until we know, not just in theory, what makes a good architecture and how much we can rely on the latest model and its insights.

Writing

  1. Non-deterministic AI and the illusion of control

    Some 20+ years ago I had my first big release, for a high-traffic news website. It went down so quickly that I'll never forget how shocked I was.

  2. For 20 years, advantage was a question of who you hired

    The edge is moving from who you hired to how your team uses the model.

  3. One of the best moments every year

    150 trainees start their careers, and the impression we hope to leave.

Press & speaking