Uncle Bob — author of Clean Code, author of Clean Architecture, the man behind the SOLID principles — doesn’t read the code his agents write.
Bob writes:
“I’m considerably older than you. I started programming in the late ’60s. My current strategy is to not read the code written by my agents. It’s the only way to actually benefit from their throughput. Instead, I surround the agents with extreme constraints: unit tests, Gherkin tests, QA procedures, quality metrics, mutation testing, test coverage, and a bunch of others. In the end, I’m very confident in the code they produce, because it had to pass through all of my constraints and tests.” https://x.com/unclebobmartin/status/2080257779395154409
And a couple of days later, he ran his usual bit — the post-sauna unpopular opinion:
“AI agents can write code many times faster than you. Maybe 20x. What are you going to do with all that extra time they’re giving you? You’re going to have those agents, with their 20x throughput, write unit tests, acceptance tests, and QA tests. Then you’re going to mutate those tests and run quality checks. You’re going to torture that code. You’re going to write property-based tests. You’re going to write performance tests. If there’s concurrent code, you’re going to write jitter tests. And you’ll have the agents do all of it, because they’re 20x faster than you. And when you’re done with all of that, you’ll have code that’s better than you could ever have produced yourself, and you’ll have it in far less time than would ever be possible for a human.” https://x.com/unclebobmartin/status/2081334541667410312
If Uncle Bob has genuinely stopped reading code, maybe the rest of us don’t need to either — got to keep pace with. So I tried to remember what exactly Bob has built lately, and it turns out he’s mostly been writing books and blog posts since 1995, and by now claims “30 years of experience” — fair enough, that’s professional experience dating back to the ’70s, so the number checks out.
Two years back, on the Staff Engineer channel, we covered Bob’s fight with Casey Muratori (the video’s gone, but a stream survives — https://www.youtube.com/watch?v=PWvcfqVBb_g). Casey went through it point by point: why Clean Code is actively harmful from a performance standpoint, and how the book has aged into something outdated, even damaging. Right around then, Bob announced a new edition of Clean Code (https://x.com/unclebobmartin/status/1820484490395005175), adding chapters on AI and LLMs, beefing up the TDD material, and folding in content lifted from his own Clean Architecture.
In an interview on July 22nd, Bob said:
“You know, I’m doing little tools, little things, little fun side projects. I’ve got one project for a nearby flight school — I help them maintain a status board, things like that. I’m not in the realm of huge systems anymore.” (11:00, https://www.youtube.com/watch?v=RxxxGkFIUJ0)
And there it is — everything clicks into place. What Bob needed to say was: “My current strategy is not to read the code written by my agents… for pet projects.”
And honestly, Bob, you don’t need to wrap those pet projects in all of it either — the unit tests, the Gherkin tests, the QA procedures, the quality metrics, the mutation testing, the coverage, and the rest. Nobody is ever going to pet project code, Bob.