Well, quite a bit has changed in the seven months since I wrote Optimize for Humans. Its key takeaway was that AI is no excuse to abandon readable code, good abstractions, and especially the people who understand how software works.

Looking back, I realize I was more attached to the idea of software development as a craft than I cared to admit. I'm no longer carefully chiseling every line of code like I did back in 2022, but I'm not team never ever look at the code either. I am still figuring out where to place the cursor, though admittedly it's been increasingly drifting toward the never-look-at-the-code end.

For years, I loved carefully writing code, paying particular attention to naming things. Nothing beat the satisfaction of making every piece fit, simplifying something until it felt right, then enjoying that blissful moment when everything finally worked.

It took me some time to let go, but I have outgrown the need to personally shape every line of code.

Technology is a means to an end, and that end is making people's lives better. AI has given me the ability to build things I previously lacked the knowledge or time to make. Enjoying that newfound freedom has changed how I spend my days and what I find satisfying about the work.

I still care about clear code and simplicity, even as I spend less time inspecting the implementation myself. I want to understand what I'm shipping, I have yet to determine how much understanding is enough. The possibility of AI becoming unavailable makes me wonder whether I could maintain what I've built without it.

Even in the areas I know well, the models now regularly suggest technical solutions better than the ones I would have come up with myself, whether it's syntax, unfamiliar libraries, or even architecture decisions. What I now spend my time on is deciding what the software should do, discussing how to get there, and evaluating the result. I'm the software engineer I always wanted to be.

I believe the path moving forward is even more delegation to AI, and more focus on outcomes. For me, that means taking on more product decisions and leading teams of agents.

I have been having a blast. There is a lingering dread about what these same capabilities mean for my livelihood, and building more hasn't made that disappear. But the pleasure of using something that existed only as an idea a few days earlier is very real, and I am getting to experience it a lot more often.