Skip to content

Reading code metrics in context

Hot take:

Judging developers by lines of code might not actually be a terrible idea.

The problem is judging them by lines of code alone.

High LOC with minimal deletions can mean someone is contributing substantial new code, but it’s not as simple as “high LOC = good work.”

High LOC paired with a high number of completed features or tickets can mean the work is moving and the code is landing.

Low LOC paired with a high number of completed features can mean extremely efficient use of tokens, attention, and implementation effort.

In a perfect world, people who do this consistently are going to be your top contributors. They communicate well, understand the process, and know how to turn intention into shipped work.

High LOC with low feature completion, heavy deletion, or constant rewrites can mean churn.

That pattern can point to unclear requirements, uneven communication, architectural confusion, or wasted time and money.

Given the current state of tooling, some of this is inevitable. What you really want to watch for is habitual behavior.

Commits and conversion

Commit hygiene matters too.

There’s a meaningful difference between a clean 250-line commit that can be reviewed, tested, and understood, and a sprawling 25,000-line commit that lands like a meteor on your codebase.

Did we really need that 15,000-line OAuth refactor?

Well, no. Probably not.

I think the nuance gets lost when people flatten productivity into being impressed by token usage.

It’s too shallow a worldview to be useful.

Instead, let’s see if we can peek into people’s brains a little.

What were they trying to do?

Where did all those tokens go?

How much of that motion became working software?

When everyone can work at a mile in a minute, the thing you have to look at is conversion.

How far does your money go?

Happy Sunday, everybody.