Andy Halliday

It uses Gemini 3 to do that feature.

There you go.

There's your intro to Gemini 3.

I mean, LOLs can do that.

Can I interrupt you just to ask you to do one quick thing?

Will you show us the prompt that you use to generate that output, which is an instruction set on how to prompt?

Because I'd love to reproduce that.

And this is a feature that has been missing in most LLMs.

If you ask it about its own capabilities, it doesn't really speak very knowledgeably about that in history.

But this is impressive insofar as it's got what amounts to a built-in tutorial for how to work best with myself.

Using those terms.

Was the term objective-based commander's intent prompting something that you learned first in this dialogue?

Okay.

It's probably in the documentation somewhere, but typically when you want to find out how best to work with one of the models, you have to go to one of those help documents that have been posted by the developers, right?

But here it's all built in.

I really love that.

So let me ask a question about that.

So there is a canvas in Gemini 3.

Did it present this on the canvas first before delivering a downloadable HTML?

everything for option C. So things that I would improve on that is canvas in the tools selector down there.