Tell us how you came about thinking about this EQM score and the work that's been done behind that.

Yeah, thank you, Chris.

You know, I think Javed said it so nicely.

We had an understanding fundamentally that these conditions coexist and they may interact with each other.

But we've had sort of serendipitous wins that I think really kind of pushed us forward.

One is that there are therapies, a singular therapy has shown benefit across multiple of these disease states, suggesting that some of these therapies may be modifying underlying pathways that exist and that are risk factors for all of these different diseases.

The second is, in some cases, we've been lucky.

Tell us how that works.

There's this inflammatory state that can be measured, and then there's this functional state.

Yeah, thanks, Chris.

I think we're still in the early innings of this story.

And as Martin and Joanne said, the field has moved incredibly quickly over a short period of time.

You know, what we know are there is that body weight obviously is an easy metric, and maybe that's why we anchored to it first and something we've routinely done.

But the distribution of fat, how metabolically active or inflammatory active it is, likely play an important role.

You know, if I could sort of zoom out and Martin and Joanne made this analogy to heart failure, if you think about heart failure with preserved ejection fraction, there's always this natural tension between developing therapies for a very large population or developing potentially more effective therapies for a more narrow population.

Ankit, how would you guide development in this area?

What are your non-negotiables when you're creating a prompt to ask something? Can you help people understand what do you think about when you're doing that prompt where you say, "If I don't do it this way, it's not gonna come out right," or, "These are the things I make sure I include such that I'm getting the right answer"? Can you give people a little bit of just actual real-world guidance?

Yeah.

Thank you.

In fact, the large language models are, are really a completely different workflow than what we've been used to.

And a few practical tips is you get out what you put in.

And so what do I mean by that? Uh, what I mean by that is you want to be quite specific in your prompt.

This isn't, you know, the way you would operate a Google where you just hit a few keywords and get sort of a large swath of maybe more relevant content.

I prioritize safety over something." What do you think about that? Because we hear a lot about that now, and so I'd love to just hear directly from you for our audience.