I don’t claim to know what the future holds for AI software. But I this post aligns most with where I see the industry going in the short term. As a senior engineer I spend a lot of my time reviewing code, questioning assumptions, and ensuring quality.

The current AI models are really good at spotting out lines of code and agents can be configured to be better at writing tests. But they fall short at holistic quality. They can refactor code and get it working but they often make a mess that becomes harder and harder to change with successive updates. This isn’t all that different from human written code. But quality needs to be something that is optimized for and validated. It feels like that is the job now. Ensure quality. You can write thousands of lines quickly and build really cool things. But who’s making sure it’ll be something that is still workable in 12 months?