Working with AI

I use Cursor and Codex most days. Not because I have found the perfect way to work with AI, but because I am curious about what it changes and I enjoy learning through the work.

The environment moves quickly. New models, tools, and ways of working keep appearing, so my process is still adapting. I try things, keep what helps, and stay careful about what I do not yet understand.

Less typing, more thinking

A lot of development used to be manual code writing, remembering syntax, moving between files, and translating an idea into the right semantics. AI can now help with much of that mechanical work.

That has moved more of my attention toward the parts I have always found most interesting: understanding the real problem, deciding how a system should behave, finding the awkward edge cases, and making tradeoffs that leave the product simpler. Those hard parts were always the juice.

The responsibility is still mine

The tools can move fast and sound certain when they are wrong. I still need to understand the change, make the product calls, test the result, and own what ships.

I do not think I have reached a final method. I hope I never treat it that way. For now, I am staying curious, doing the work, and learning how to use the extra leverage without giving up judgment.

The code may take less time to write. The interesting work is still figuring out what is worth building and how it should work.