The AI Situation in Software Development

Working with AI in coding shifts your time from implementation to upfront design and detailed prompting. You can give a full spec, a high-level goal, or something in between, but each approach costs time. Context windows limit how much you can feed, so you must compress and refine. Domain-specific models matter, and feedback loops with tests and tooling are crucial. Ultimately, you still spend time, but now it's in words, not code.