Software Has Changed

A tweet from Jamon Holmgren has been making the rounds this week, and it's the kind of thing that makes engineers stop scrolling and nod. His point: talk to any experienced developer one-on-one, off the record, and you'll find people quietly shell-shocked by how fast their job has mutated in the last year or two.

The shift isn't subtle. Code gets written and discarded constantly now, because generating an experiment costs almost nothing. That old feeling of flow, hours locked into a single problem, has been replaced by something stranger: sitting there watching a model think, waiting for output, then judging it. Holmgren says output has gone up, but he's careful not to call that progress on its own. More code isn't automatically more value. Faster isn't automatically better.

Where he lands, though, is more optimistic than the doom-scroll version of this story. Because devs aren't burning hours typing out every line by hand anymore, that saved energy goes somewhere. Reviewing more carefully. Refactoring instead of shipping and forgetting. Catching technical debt before it ever merges to main, rather than paying interest on it for the next three years. In his framing, the engineers who were already good get to be better, since the busywork got automated away and the judgment work is what's left.

He's blunt that this isn't a world he asked for. Software engineering changed under his feet, essentially overnight, and he's still working out how to feel about that. But he's chosen to lean into thinking at a higher level instead of grinding at the line-by-line level, which he says helps him avoid tunnel vision on any single piece of code.

His closing question is the one worth sitting with: if the final code is solid, thoroughly reviewed, and something you'd actually trust in production, does it matter whether a human typed every character or an AI drafted most of it first? That's not a rhetorical shrug. It's basically the new job description.