Am I the only one? (A reflection an AI coding agents)

This past week has been indistinguishable from any other, at least in how often my AI doubt has been triggered.

A colleague tells me he now “just lets Copilot write everything” while he leans back and supervises.
At a christening, a friend casually predicts he’ll be out of a job within a year because the models already produce architecture with “better taste” than he does.
My uncle proudly demos something he built with Loveable and asks which model I prefer; ChatGPT, Le Chat, Gemini? As if the choice were now a basic personal setting, like a browser.
Even my carefully curated YouTube algorithm keeps insisting that “coding is solved” and the only remaining skill is discovered in tutorials on orchestrating agents. Even my last hope, Dave Farley, seems converted to some extent (and I know YouTube content creators are also affected by what generates clicks).

I have this uncomfortable feeling that I’m missing something important. That I am being contrarian out of pride rather than logic. That I am being lazy because I won’t change my habits. I mean, how can so many people agree on something I just don’t understand?

Let me be clear: I genuinely want the agent approach to work! I have too many things I want to build and not enough time, and although I enjoy programming, I enjoy having the finished system more.
That is why I have installed and removed Cursor from my laptop at least five times. I also tried Claude Code and Copilot. But the cycle stays the same: I prompt the problem and describe what I want, watch the agent do a lot of impressive things, and then stare at the result wondering, is this actually what I meant? And is what I meant what I want? And after a couple of attempts at the grand “solve it all up front and fast” approach, I always return to my slow, small-step, test-first development cycle. AI coding agents, once again, back on the shelf.

The irony is that I love AI. I love asking ChatGPT for recipe ideas, asking Claude Code for design or algorithm suggestions, and hiding my limited English vocabulary behind a prompt or two. I don’t even feel ashamed, although I did remove the double dash from this post so the world wouldn’t know an LLM helped me sound intelligent (wink wink). AI has truly changed my life, but mostly through inspiration and significantly better autocompletion. Here is what has not changed for me: I still code a lot.

For me, coding is a tool for problem solving. And problem solving is really about analysis and deep comprehension. To tackle complex problems, which we will undoubtedly face more of as time goes on, we need to narrow and decompose them. Each decomposition, or each solution to a decomposition, can lead to discoveries that change the problem itself. That’s why, for decades, software development has been about shrinking the iteration cycle: making each loop smaller, faster, and more testable.

AI coding agents can certainly handle small steps, but they do so as assistants prompted with an initial task. The adaptive judgment needed for true problem solving, deciding how discoveries reshape the next step, still requires a human in the loop at every iteration. In other words, agents can accelerate the mechanics of coding, but they can’t replace the core iterative thinking that drives meaningful software development.

So what is my point?

I’m not against AI. I love it, and I use it every day. But I use it almost exclusively for very local, very specific, context-bound problems. I might ask it to autocomplete a function, suggest a small algorithm tweak, identify a memory leak, or even write a comment (yes I said it) to my function signature. Great software is built by taking thousands of small and sharp decisions. If an AI agent is working for you, that’s great, but I highly doubt you are creating anything meaningful.

So here I am. Still copy/pasting my code into an online LLM asking, “Finish my function.” Still getting comments like, “You probably need to practice your prompting.” Still using NeoVim like I’m living a decade behind. And still feeling that uncomfortable sense that I’m the dinosaur who will be left behind. Am I the only one?