One year ago, I would ask Gemini or ChatGPT to help me with some small scaffolding issues. Help me debug an issue, and it was a 50/50 success ratio.
Now, with Claude Code I co-develop features, I can steer it into a specific way of working, and it knows my preferences. A major improvement. Still not perfect though, without clear instructions it will build something of itself and often misses. But, what it tells me is that if you can provide it clear requirements and instructions how a solution should fit in your architecture, it can build a good enough solution rather quick.
But that is not the end! It is the beginning. With this ‘make it work’ phase, you can now look at your diff and refine further. As long as you don’t make super big PR’s (which you never should anyway), this is a workable process and you can get things done more quickly than before.
The downside? I feel like it is harder get ‘deeper’ into the problem domain, and fully grasp the edge-cases. AI can help with that, but it also hallucinates and if you’re not deep into the subject matter you will not see it. (you don’t know what you don’t know).
For now, this is where I’m at. I see value in my daily work and I see how AI can help me.
Alongside programming, I see also AI agents help with reviewing. And in fact, combine agents with their tasks you can quickly resolve feedback and only deal with ‘real issues’. I also recognise a certain brain fatigue. You need to quickly grasp concepts, big chunks of changes and switch context more often. I don’t know if that is good for the quality of things. As with every new “hammer” the answer would be to throw more “AI” to fix these problems.
Who knows, at some point it is. I definitely see AI getting better and better. But it is not at the level yet you can fully trust it. This begs the question how the industry will evolve.
As a little hook to my previous post: I am collecting data and thoughts about the major (or lack of) lay-offs and how AI is supposedly changing the world. There are changes for sure, but I want to keep it nuanced. So stay tuned for posts about that.
With all that being said and done. I do feel a little sad. For me, programming, the typing, thinking, everything, was a major joy for me. With AI a lot of that joy is gone. I wonder how that is for other developers out there. What are your experiences so far? Is AI helping you? Do you enjoy it? How about the things I mention above? Recognisable?
I’d love to hear your thoughts!