The Keystroke Is Not the Measure

There was a half-joke about a metric once. CPM. Characters per minute. How fast you typed. Nobody put it on paper, but we still ranked developers partly by throughput -- how quickly they produced, how much they got out.

Now AI pours out the code. And suddenly it becomes visible how meaningless that metric always was. It does not matter whether you type thirty characters a minute or a hundred and twenty. The model does it. The keyboard is no longer the bottleneck, and therefore no longer the measure.

What Is Left to Measure

If not the production pace, then what?

The question. The direction. The Socratic nudge that gets the model going the right way. And above all: the ability to look at what came out and know whether it is right. A senior who asks the sharp question and validates the result is in an entirely different class from someone who just feeds the tool and hopes. Writing was never the point. It is just obvious now that it wasn't.

The Trap We Haven't Let Go

Despite that, we still measure half and half by volume. Lines. Commits. I have written about how easy it is to be fooled by it, and I still catch myself reading a list and thinking that the one who typed the most did the most.

But one of the most senior people I know sits down one day, thinks, checks in four lines, and goes straight to production. Throughput is a vanity metric. It measures movement, not value.

What the Proxy Hid

The keystroke was always a proxy for thinking. We measured the fingers because we could not measure the judgment. AI removed the significance of the fingers and left judgment exposed, with nothing to hide behind.

It is uncomfortable for those who used writing as proof that they were working. And it is good news for those who always had the judgment but never the fastest fingers.

See also: My Job Is the Questions and Give me the insight, not the decision (series 34).