How my job changed in a year of AI

A little over a year ago AI was a clever chatbot I asked the odd question. Now it runs most of my day.

Remember Stack Overflow?

Early 2025, AI was the thing you turned to instead of trawling Stack Overflow. A bug, a stack trace, a “why won't this compile”. Paste it in, get an answer, no closing three cookie banners on the way. I started using it for personal things too: gym plans, diets, the sort of question a Google search technically answers but only after four articles and a wall of adverts. Back then the workflow was crude. You copied an answer out of a chat window and pasted it into your code by hand.

Then it moved into the repo

The real shift was AI getting into the codebase itself. Suddenly it had context. It understood the project, the patterns, my coding style, which meant it got things 90% of the way there, fast. That was brilliant for the fiddly stuff, the picture-perfect React work that used to eat an afternoon. Now I can hand it a screenshot and say “build this design,” and it largely does.

The bit that sounds bonkers

Here's where I lose the non-technical people. A feature now moves through a pipeline that's mostly AI:

AI writes the ticket (a prompt lives inside it)
        |
        v
AI picks it up and does the work
        |
        v
AI reviews its own work
        |
        v
Opens a pull request
        |
        v
A different AI reviews it
        |
        v
     Merged

We're seriously toying with an out-of-hours agent that picks up work while we sleep. Say that out loud and it sounds like I'm automating myself out of a job. Maybe I am.

We use Claude Code for basically everything at work. I keep ChatGPT around for personal use, but I don't reach for it for the job any more.

Is it actually faster? Yes, used properly

I've read the pieces arguing AI doesn't really make you faster. I disagree, with a caveat: used correctly. The pace at Covertime is genuinely mad: we ship real features every week, and production issues have stayed minimal. That's the honest headline.

The overhype is real too. People now use AI as the first answer instead of trying anything else, reaching for it to name a variable or recall something they already know. It's a tool, not a reflex. I'm glad I've got a decade of skills sitting underneath as a fallback, because when the output is confidently wrong (and it is, often) you need to know enough to catch it.

I feel for the juniors

I'm going to become the relic who says “back in my day we wrote it by hand.” We did. It was writing, debugging, and a genuinely mad amount of time spent just naming things.

My teachers used to warn me I wouldn't always have a calculator in my pocket. I have several now. I suspect “will engineers even need to know code?” ages about as well as that line did, but not in the way the doom pieces suggest. The job changes; it doesn't vanish. Juniors entering now will build a different skillset to mine, not an absent one, closer to directing and verifying than typing. What worries me is that the way I actually learned, breaking and fixing things by hand for years until it clicked, is the bit we're automating away. I don't have a tidy answer for that.

AI is here to stay, in one form or another. I love working this way. I just hope the next lot still get to build the same foundation I now quietly lean on.