I’ve been working in software development for a whilie now and have consistently chosen to stay in an individual contributer role because, among other reasons, I love working and learning together in a team solving problems with code.
So when I say, with complete honesty, that I have never enjoyed developing software more than the last six months or so while developing together with large language model agents, it’s not because I haven’t enjoyed software development in the past. I think it’s mostly because having an incredibly fast and knowledgeable coding assistant allows me to stay in the flow of the actual creative process (yes, for those unfamiliar with writing software, it can be very creative and fun!) - designing, learning and architecting the solution - rather than constantly deep-diving into the depths of some library or framework to solve some small blocker.
But the benefit of having an incredibly knowledgable, if some-what over eager, coding assistant comes with a lot of dangers and pitfalls as well. The same capabilities that make these tools powerful - their speed and broad knowledge - can tempt us to skip the learning process and generate code that we don’t fully understand, creating downstream issues for review and maintenance.
In this post, I want to highlight the benefits as well as the strategies to avoid the pitfalls, in a top-10 tips format. Hopefully it’s helpful whether you’ve thirty years experience or three.
