These blog posts, lamenting losing something, create fascinating juxtaposition against posts from people having more fun than ever, and they’re often published on the same day.
Hot take:
Coding by typing sucked.
Being stuck on some esoteric problem and having to rubber duck your way out of it. Your value being connected to the amount of stuff you could do that few others had the patience for, instead of how empathetic you could be to user needs.
Even wilder, these laments ignore history.
In “The Mythical Man-month” they talk about spending weeks debugging wiring on a breadboard so a chip could be designed based on said wiring. Does anyone in 2026 wonder if their CPU neds debugging? Only the most esoteric of situations and it’s only a last resort. I’ll concede it’s not zero people but it’s not more than a handful. A faulty chip design is the last thing you think of when you have a bug.
In the 80s there was a “golden era” of coding 8-bit machines where you could understand the whole machine. Did anyone worry their CPU was mis-executing instructions? No, their idea of “understanding the whole machine” depended a huge pile of abstractions that had been made reliable thanks to solid engineering.
Can you understand the whole machine in 2026? How it boots, how your filesystem works, where the caches are, all of it? No! There was a point at which we let go of that need in order to gain greater leverage.
At some point you have to let go.
Current coding agents are in their “debugging breadboard” era - they’re unreliable and not worth the valuations the companies are getting. The craft is changing though, and letting go is the future.
The challenge we face now is how to engineer probabilistic systems into shape. But guess what? Most other engineering disciplines deal in a messy thing full of probabilities called reality. We have tools to deal with this - tolerances, fudge factors, the works.
Complaining about progress is going to age really poorly.