Context is the brief

More real-world examples of why vague AI briefs still bite

I spoke about the importance of being precise and thorough when briefing AI.
Naturally, I caught myself missing that mark twice, today.

First: I had Cody in a multi-app setup and almost let it proceed after asking it to “remove the bottom section from the dashboard.”

That sounded specific until I realized it started working on the wrong app, the wrong dashboard, the wrong section.

Reversible, sure. Still scary (and dangerous!)

Second: After changing some OpenClaw skills yesterday, I noticed it had gone weirdly quiet today, ie. no daily curated news.

So I sent:

havent seen the news today

Zero context, like I forgot it’s still running “ChatGPT” under the hood.

The response:

It’s technically a good answer to my question

I guess, sometimes, the shoemaker’s children go barefoot …