This morning’s episode of The Daily featured the reporter Clive Thompson discussing an article he wrote in mid-March, titled Coding After Coders: The End of Computer Programming as We Know It. The discussion about the article was great in its own right, but I’d encourage anyone who heard it to consider it something of a teaser for the article itself.
Thompson’s article was one of the best things I’ve read about the way software engineers like myself are using AI on the job these days. He’s not exaggerating. I barely write code by hand now — an AI agent writes it for me, quicker by far than I would be capable of. I lay out a plan for the agent, define its success criteria and the ways in which I want it to code, and then correct the agent if it goes off track. With time in any given codebase the velocity of the agent increases as I help it get more and more efficient at understanding exactly what I’m expecting of it.
Let me be explicit about the change that’s happened here. My whole career I have had to build software by writing very precise instructions to the computer. Now I build software by having an ongoing series of discussions (actual conversations!) for weeks with AI agents, and those discussions result in both new software, and AI that gets faster and more precise at building that software. How wild is it that not only does this work, but it results in software being completed anywhere from 2 - 10 times faster than I could build it alone?
This is heady stuff, and while it’s a change that might be starting with software developers this will be filtering down through the entire knowledge economy. No highly technical white collar job is going to look the same in five years.
A couple other solid AI-focused articles in the Times recently that are worth your time:
- The AI Disruption We’ve Been Waiting For Has Arrived, by Paul Ford
- The Big Bang: A.I. Has Created a Code Overload, by Mike Isaac and Erin Griffith

