Cognitive offloading vs cognitive surrender: How software engineers should actually use AI

追加: 2026-05-23

[00:00:00]There was a lot of concepts that felt very familiar and one of them was like the human in the loop aspect to it. I had not I had not seen something that was so useful to interact with a human built in in a distributed system other than something like Spinnaker. And and it's interesting cuz like Spinnaker, you know, like continuous deployment and everything like Kubernetes. Then I actually ended up reading about how Netflix replaced the engine running behind Spinnaker to be Temporal. Uh which was really interesting. Like it actually makes a lot of sense that kind of like a new piece of technology, at least new for me, that fits very well with the existing environment. And it does make a lot of sense for people running AI because like, you know, like if anything AI is probably random.

[00:00:45]>> [laughter] >> Well, you don't know what's going to happen and what things are going to happen. So, having that ability of not only replaying when things go badly, but also being able to inspect when things go weird to understand what happened sounds very useful, yeah. Um what how do you feel about what's going on with AI right now? Like I know you've played around with building with AI, so I'm curious where what's your opinions?

[00:01:08]They're two different things, right?

[00:01:10]Building for AI and using AI. Yeah, so I'm also a musician and I feel like AI when it comes to creative areas is a little bit not necessarily problematic, but it can be problematic for certain usages. When it comes to specifically us software engineers, I feel like it is really powerful as a as a way to enhance the way you think. Actually, I read I read an article recently that talks about the difference in between cognitive offloading and cognitive surrender, right? So, like the difference in between hey, do this for me and then I'm going to verify that the agent did it correctly and that it matched my expectations, etc. versus cognitive surrendering, which is like, hey, do this for me and whatever the output is, it is my output, right? Like, you In In the first case, you're actually using AI as a tool to prove your thought process, and it's great. I think that actually enhances your ability to produce better things. In the second one, though, it feels like you're kind of replacing yourself over time, and eventually you won't be able to understand your own code bases, right? So, like, it's There's a trade-off in there, for sure.