“Coding has never been a bottleneck, and I tend to agree with this because building the right things has always been the bottleneck, and it's a bigger bottleneck today than before because of the speed of coding. … We can build faster, but are we actually building the right things? If we are not, it doesn't matter how fast we can move.”
sputnikintelligence.com
The host asked how staff and principal engineers should cope with a flood of AI-generated pull requests. Ojstersek said to slow down, and that the real constraint was never typing code but choosing what to build.
…skills you need a lot more you know managing company politics uh and being good kind of like in having building good relationships with people outside of the engineering with other departments and so on but i think that that is the way to go these days i would that that's my recommendation
…we're building good products and i feel like now you've got these staff principal engineers who not only have to be people please you know people pleasers on both sides to above them and below them but now they're trying to manage this new world of like well code is cheap and we're getting a lot of
it yeah it's it's a really really good question i think there's a bunch of uh insights we can mention here especially you know the first thing is what i recommend is slow down instead of just trying to rush all the time because i think these days um we are all expected to make miracles with with ai but the problem is with uh when moving fast it has never worked out well in any in any of our eras in engineering that that's that that's my understanding is if you're rushed to do things you're always going to to generate slop and bad decisions and so on of course yeah we can fix them along along the way but the problem is um if if if you deliver slop you need to have a good trade-off in mind right and there needs to be a good business decision there needs to be like a really critical thing that we are going to just quickly deliver it and then we're going to make it better but if if it's not i mean it's it's not an excuse because speed of coding is so fast that we can just deliver slop and that's also like whenever i see a pr from an engineer and i see a bunch of ai slop in there and then i see a bunch of ai comments in…
Yeah, I would agree with you. I've noticed that slowing down to speed up helps. Like just people take, just like taking 10 minutes, you're starting a project, just thinking to yourself, just thinking by yourself, even if you want to use the LLM to help you research, I think, or think, instead of…
Yeah, and I mean, thanks a lot for having me, Daniel. I really enjoyed this conversation. I think, you know, it's all about being a continuous learner. Don't get too much into trying to learn every new shiny AI tool. Fundamentals are more important than ever. You know, also, yeah, as we talked…
Awesome. Thanks for your time. I appreciate it. I feel like this is a masterclass and so worth it. Definitely. Everybody go check out the book, buy it on Amazon. I appreciate your time today.