#ai

To prevent code drift you need to make AI take small, well-defined steps.

AI can accelerate development, but it doesn’t remove responsibility to understand the code.

Understanding the illusion of intelligence in large language models.

Where did “don’t execute untrusted code” go?

We try to bottle architectural intent, but the world’s messiness always seems to leak out.

If you are not in a position to judge the result, consider not using AI, or ask for help.

If you use AI to solve puzzles for you, you are kind of missing the point.

Don’t use AI to write text for you, or you’ll lose the opportunity to sharpen your thinking.

Generative AI is bad at coming up with novel ideas, but really good remixing existing.

What will you do when the AI abstraction inevitably leaks?

Almost by definition, a suggestion by an LLM represents the least surprising solution.