Up to this point, I’ve mostly used AI as a sounding board and as a more-precise Google. I’ve talked through high-level architectural challenges, asked about forgotten syntax and implementation details, and had it check my logic on design decisions. For that sort of thing, I can’t deny it’s been extremely useful. Almost magically so.

What I haven’t used – and what I’ve been avoiding – are coding agents. I know they’re taking over the industry (and burning through token budgets in the process), but I’ve been resistant to try them. My complicated feelings are probably worth examining at some point, but that’s not what I want to do today.

I might be the last developer to realize this, but AI is here to stay. And as much as I loved the old way of coding, it’s now in the past. At least in part. If I want to stay sharp in this field, I need to accept that and become familiar with these new tools.

So here’s my experiment: can Claude Code be more than fancy Google and a replacement for my rubber duck? Can it take on chunks of implementation in a useful way? That’s one facet of the question I’m trying to answer, but it goes deeper: can I successfully integrate it in a way that I’m happy with? Can it become a real tool and not one I’m forced to use? I don’t know the answer yet, but I intend to figure it out.