Andy Halliday

So for example, thanks for this guidance because I haven't used it yet, in a doc.

It has a writing editor, so you can click on that, and that gem will then kind of give you advice on improving the form of your document.

Brainstormer, copy creator, sales pitch ideator.

So they have a nice suite of gems.

You don't have to go build your own to get some utility out of using this within Google Docs.

Oh, I want to just talk about it being December and we're going to start thinking retrospectively about what happened in 2025 and what's going to be happening in 2026.

But I just wanted to say what a difference a year makes.

Like I'm going to talk about this very large state of the AI space.

base that was just published by Open Router.

And they're in a unique position because Open Router, of course, sees all the traffic to all the different models that are, you know, by the users of Open Router.

But anyway, they just published this very long, I think it's 70 plus pages, document that I'm going to feed to Notebook LM shortly.

But I only saw it for the first time this morning.

It came across the news wires.

And it's primarily about 2025 being not necessarily the year of agents, although agents are important that way, but it is the seminal year during which reasoning models came to the fore.

And

that the field of AI shifted inference from a single pass sort of prediction response, right?

So the dense model spits out the response to whatever your prompt was to multi-step deliberation inference.

So if you rewind the clock, a little over a year ago, we started to hear from OpenAI about some model they had called Strawberry.

And Strawberry was inside and it was doing something peculiar that the other models hadn't been doing before.

Not peculiar.