Engineers should manage agents, not blindly ship their code

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 11:39

   So one of my tweets that went viral the other day, I called it reject non-understanding. So what do I mean by rejecting non-understanding? So there is two schools of thought today. One, you can tell your AI agent to do whatever, and you don't even look at what it's doing. And you press a button, ship it, and it ships a supercell or whatever. And there's another route, which is the one that I recommend to my children.

2. 12:08

   literally I was up doing math homework with one of my children last night. It's actually use the AI, but also understand what it's doing. Now, you asked specifically, do I write the code? I read code. And what this says, I'm doing more and more and more is asking the AI to explain back what it did to me, which is a little bit of the job. I've compared it to being an engineering manager. The engineering manager at any big tech firm in Silicon Valley over the last 10, 20 years, it's also been a subject of debate.

3. 12:42

   But for the most part, they help their engineers get organized. They help them execute, get blockers out of the way. Obviously, the best engineering managers are amazing engineers themselves. Occasionally, they write code. Sometimes they write code to test out what you did and poke holes. Sometimes they might write a little benchmark. Sometimes they might do QA of the final product. Sometimes it might bring, you know, in collaboration with the product manager, it might bring feedback from the field.

4. 13:08

   So they're kind of looking out for you to not make mistakes. And so I think the role that most people are now taking with this coding agency is that of an engineering manager. But again, the best engineering managers that I've ever worked with, you know, first of all, they love code. They love like writing it, understanding it, et cetera. And so I still do it. I also shipped this weekend, I shipped something you could call a proof system, which is a mechanism by which we can make sure that the AIs don't ship slop and don't yeet your data out into the world by mistake.

5. 13:42

   And I purposefully read lots of the code. I wrote down a lot of the examples, et cetera, because there's a lot of sensitivity around code, right? Like AI and code will run more and more and more of our lives, whether we like it or not. You know, Formula One, I was just reading some news. It was this massive event because they had to ship a software update into the cars and the drivers were complaining.

6. 14:09

   And these moments make you realize, holy shit, these Formula One cars are being run by code. How the engine works is determined by code. You're going to go to the hospital and all of those are run by code. And so it's so, so important that we take this, you know, AI safety issues and the quality of the code we produce and how we verify it. We need to take it extremely seriously.