When was the last time you heard anyone talk about a 10x engineer?
The concept had its moment. The argument was that some engineers were ten times more productive than the average, and that finding and keeping them was the real hiring problem. It was contested from the start: productivity is hard to measure, context matters, and the lone genius framing ignored the system around any individual contributor. But the intuition was real. Some people produce dramatically more than others, and the gap is not explained by hours worked.
The concept faded. I think it faded because it stopped being useful, not because the underlying observation was wrong.
AI has made it relevant again, but the question has changed. It is no longer whether some engineers are 10x more productive than others. It is what happens to that gap when everyone gets a multiplier.
Two arguments exist and I do not think either is fully right. The first is democratisation: AI closes the gap because it gives more capability to engineers who previously had less. The floor rises. The 2x engineer with AI now outperforms the 10x engineer without it. The second is amplification: the gap widens because the engineers who were already strong are better at directing AI, better at catching its errors, better at knowing when not to trust it. The multiplier is larger for the people who were already ahead.
What I am seeing, from my own work and from the replies to last week's post, is that the amplification argument is closer to right, but it is measuring the wrong thing. The gap is not widening on the old axis. It is shifting to a new one.
The 10x framing was about output: lines of code, features shipped, bugs fixed. AI compresses all of that. The new axis is judgement: knowing what to build, how to specify it precisely enough that the agent can execute it correctly, how to verify that it did. The engineers who are pulling ahead are not the ones who type faster or know more syntax. They are the ones who think clearly about the problem before they touch the tools.
A 10x engineer proficient with AI is not producing 10x the code. They are producing 10x the clarity. And clarity, it turns out, is what was always scarce. Murderbot hacked its governor module not to stop working but to work on its own terms. The judgment stayed human. The capability expanded. That is the augmentation that matters.