Context Over Prompts

Most people spend their time trying to write better prompts. I think they're solving the wrong problem.

A prompt is only as good as the context surrounding it. An agent with a mediocre prompt and excellent context will often outperform an agent with a brilliant prompt and almost no understanding of the project. The long-term opportunity isn't prompt engineering. It's building systems that continuously preserve, organize, and expose context so neither people nor agents have to keep reconstructing the same understanding from scratch.