How AI Changed the Way I Write Code

AI has completely transformed my daily coding workflow. Instead of starting from a blank file and manually designing every detail, I now treat AI as a smart collaborator that helps me plan, implement, and refine features much faster than before.

From Code Writer to Code Editor

These days, I rarely jump straight into writing code by hand.

I usually start in a planning mode, where I talk to the AI about what I want to build, the constraints, edge cases, and how the solution should fit into the existing system. Together, we shape the approach: I guide it, challenge ideas, refine the plan, and only once I’m satisfied do I ask for an implementation.

For smaller tasks, AI often writes most of the code. My role then shifts to reviewing, editing, and integrating it into the codebase. This makes me less of a pure code writer and more of a code editor, focused on correctness, design, and maintainability rather than typing speed.

Letting AI Think About the Task

For many small and medium-sized tasks, I no longer need to think through every implementation detail myself. I describe the goal, inputs, outputs, and important constraints, and let the AI propose a solid solution.

I still think at the system level—how the code fits into a larger feature, how it interacts with other services, and what could break later. For bigger features, I spend more time upfront on architecture and planning, but once that’s clear, AI still handles most of the implementation.

AI excels at boilerplate, repetitive patterns, and exploring alternative solutions, which lets me preserve mental energy for higher-level design and problem-solving.

Reviewing AI-Generated Code

Even though AI writes a lot of the code, I never treat it as ready to merge.

I review AI-generated code the same way I would review a junior developer’s or intern’s pull request: does it follow conventions, is it readable, and does it match the original requirements?

I check for best practices, internal guidelines, security concerns, and performance issues. Tests are run or added when needed. Linters, formatters, and CI checks help catch simple issues, but responsibility for quality still rests with me.

AI accelerates implementation, but final accountability remains human—and that balance feels exactly right.