Agents make it very easy to outsource thinking. If I blindly accept that, though, I may end up with a shallow understanding of the system.

Atomic Object is a software consultancy specializing in custom software development and product design. Readers can learn about software development methodologies, user-centered design principles, and agile practices. With case studies, blog posts, and technical insights, Atomic Object provides  guidance and expertise for building innovative and user-friendly software products.

Atomic Spin

Agentic development tools can accelerate software work, but they also risk creating shallow understanding if developers passively accept generated output. To counter this, the author sketches their own plan before prompting an agent, then compares approaches to sharpen architectural thinking. During code review, they go beyond surface-level checks — examining actual queries, network behavior, failure cases, and performance implications. This deliberate engagement has grown their engineering instincts over time, with the core philosophy being that AI should amplify expertise rather than substitute for it.

How I Use Agents Without Stopping My Own Growth

Interesting approach. I would also add layer where you do not know code ai generates ask for explanation in simple terms after that ask him to give you some tasks for beginners. Solve the tasks and improve the coding understanding.

This is the near ideal scenario when working with agents. Sadly, most developers blindly accepts changes without even batting an eye. This behavior can be associated with the responses being too long or just the general AI token usage pressure itself (which is a worse metric to track performance by the way).

If agents would be used as helper tools instead of main development tool, anyone can use it to grow

This is a question I have personally been thinking about. I appreciate that there are others thinking about the same, and trying to address it. Personally, given am not so experienced, I am trying to find ways to allow my development growth, even when I use agents for work - and this article has some good suggestions I should try out.

It’s more of this that will save us from the over marketted hype that AI currently is, and the cliff it’s leading us to. Great job!