I feel lost on AI

I’m a 46-year-old programmer from Mexico. I’ve been doing this work for a little more than twenty-six years now, and I don’t want to stop. Writing code is not just a way of living for me. It’s what drives me most of the day. It’s the first thing in my mind when I wake up in the morning, and the last when I turn off for the night. When I’m done, I hope I can be at peace with myself, knowing I did what I could to write all the code I dreamed of.

I’m telling you this, so you understand what I’m about to confess: I’m feeling lost with all this AI-coding-for-you stuff that has been going on for the last couple of years.

I don’t know how to feel about it.

I don’t know what to say when some people I know tell me that, these days, they don’t have to write a single line of code for a project, and they are happy about it. I’m happy for them, but I don’t know how to feel about myself.

Don’t get me wrong. I love Copilot. I believe it is the best tool in your coding belt. It is, in fact, a code-completion tool, but a smart one. A really useful one. I feel more productive these days. I can write more code during the day, and that makes me happy. But the whole idea of letting the Agent do all the work for you baffles me.

I don’t want to be that old guy who wishes time would stop, and we all did things the old, good ways. So I’m digging the docs and learning to use this technology. God, just saying “technology” makes me feel old.

I’ve used ChatGPT and Claude extensively. I’ve also been developing a project using Amazon’s Kiro for the last two months, and just recently started using Claude Code to make some changes to a customer's web app. I want to be good at it. I want it to work, but the thing is, it doesn’t make me feel a thing.

When I’m coding, I feel like I’m playing 5D chess. When I use these agents, I feel dumb. I feel like I’m cheating.

In the past, I had employees helping me with the job, but these days I’m running Han Solo. Mainly because the money hasn’t been good the last couple of years, I couldn't afford to pay for help. So I thought I could use these agents to run multiple projects at once instead of hiring people, because it is much cheaper, but the thing is, they're not exactly autonomous.

You usually ask for something, they think, get to work, and deliver something that is not exactly what you asked. You blame your AI prompting skills and try again, wait some more, test, ask for more modifications, and so on, until you get something similar to what you were expecting to get. It takes a lot of time and effort. It’s a skill issue, I get it, but I’m left with the sensation that I should: a) improve my prompting skills, or b) code it myself and get it done quickly.

This week I got an idea, so I did the following thing: I worked on two projects at the same time. On one screen, I had Visual Studio open, and I coded the solution with just Copilot's help. On the other screen, I had Claude Code working on a different project and prompted my way to the changes I had to implement.

In a way, working with Claude Code made me feel the same as a few years ago, when I hired junior programmers for the first time, but this time with a big difference —I had to admit: instead of waiting a week for a solution, I had it in a couple of hours. Is this the way I should be using AI Agents going forward?

Anyway, I might be getting old and afraid, but I wonder if somebody else feels the same way.

This is my first article on the site. I’m sorry if this is sad. I tried to warn you with the title. Leave a comment if you feel the same or want to share your perspective on the matter.