After being in deep in the agentic world these last few months, and having been at my current role the last 4 years, coming back to looking for a job has been an adjustment. Doing interviews is one thing, talking about my technical background was a matter of collecting my experience as a set of cohesive thoughts or notes that I could convey. But where things didn’t go smoothly, and to be fair I did expect this, was actually the manual coding aspect.
It wasn’t that my technical knowledge had eroded; I fully understood what I was doing like I always have, but it was more just clumsiness — the muscle memory of writing code by hand had atrophied and so when I asked those muscles to work again they weren’t as ready to do the work as they once were.
Even doing a simple ternary in React/Typescript, which I have done thousands of times before, didn’t come as elegantly as I’d hoped. I got there eventually but it wasn’t pretty; even though I had a good grasp of the language, an understanding of the constraints, it was just messy and slow.
That didn’t go well, and I didn’t take it well, especially given how familiar with the ecosystem I am. I was determined to do better next time.
It was clear that I needed to be doing more manual coding. I do think agentic coding is great and genuinely helpful, but it’s also really easy to fall in the trap of relying too heavily on it. I need to take a step back now and then — to check whether I’ve fallen into bad habits with it, and to try different ways of working.
Maybe the tool that was part of the problem can also be part of the solution? I needed to practice my manual coding skills to land the right role, but I also think it’s useful to keep my skills sharp: providing better code reviews, contributing alongside agents, and continuing to understand their output.
To do that, I used LLMs in the following ways to help:
- Producing coding exercises that I could complete, to simulate coding interviews. Initially as full projects I could work on in an editor, but also inline in an agent/chat session.
- Producing projects where I ask the agent to leave parts for me to complete, so I can practice working on actual projects.
- Working as a “virtual flashcard” where it quizzes me on a topic, provides feedback, generates examples, explains concepts when I get them wrong, and then moves me on to the next question.
Those were all really helpful, and they sharpened my skills. They prepared me not just for landing a role, but for being a more engaged and motivated human in the workplace — instead of one who was switching off because I had surrendered too much thinking to the agents.