Simon Willison noted an AI related neologism “context engineering”, emerging on The Socials (TM).
The term context engineering has recently started to gain traction as a better alternative to prompt engineering. I like it. I think this one may have sticking power.
I used to scoff at the term prompt engineering. Its early connotation was one of messing around with text prompts in an ad hoc fashion and seeing what happens. Adding goofy stuff like “my grandmother’s life depends on you getting this answer correct.” Not much engineering involved.
Given the explosion in the maximum size of context windows and how they’re constructed impacts LLM performance, I’m now in the camp of prompt engineering is serious business. In my head, I was noodling with a term “context programming” to express the many ways that crafting contexts for LLMs have arisen in engineering practice. C. f. DSPy. There’s also a bit of science emerging around where and how models interact with large contexts.
If “context engineering” catches on, I’m good with it.