As a researcher, my mind is always 20-40 years in the future, not 2-5. I saw all of this LLM work 15 years ago, and most of us in research knew this was coming. So I've had quite a bit of time to think about it's implications.

The last time I saw a real revolution was the 1990's — the web truly restructured society (for both better and worse). LLM-drive code synthesis will restructure a subtask of software engineering.