Christopher Monnier
speaker
197 appearances
1 recordings
1 series
first heard Jul 2026
last heard 20 Jul
Christopher Monnier’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.
Appearances
The LMs are just an ad just using averages of like, I don't know, judgment, like okay, good.
I would say more charitably, like, that's a good place to start.
You definitely want to make sure your your product is like
average level or better, but how how do you get better than how do you hill climb beyond the average?
Uh and to what Chuck was just saying, I think you need like this full suite of evals.
You need the the LLM judge to get you up to that average like baseline of quality.
Um, but then layering in like the third party experts for whatever the domain is, if it's a particular product, you'd wanna you'd wanna, you know, if it's if it's something that uh product that makes, I don't know, generates user interfaces, like a Figma type competitor.
Uh you'd you'd want you'd probably want some like UX designers to like be the ones to evaluate it and be like, yeah, that's actually good or that's actually bad.
And then of course you'd want the users.
themselves ca to to sort of complement and fill out the whole suite of of evals because an LLM, of course, and even a third-party human expert can't
ever know whether the product is like delivering value and utility to like real people.
So you need this full compliment of evals because there's no there's no way to know whether users themselves are getting value out of it.
And at the end of the day, it's the users who will decide the fate of the product, whether it gets adopted or not.
And I think I just can't imagine developing a product without having that that feedback loop in.
I think what's been cool is to see the adoption of the virality maybe of this particular method within within first within our little org and then expanding out to like other orgs within Microsoft and maybe even beyond Microsoft.
So what's cool now is it's sort of
come back to we're continuing to do these evals on um like the conversations and chats people are having.
Now we're we're expanding it to
to move along as we were just saying, like as AI capabilities evolve.
Now with many different AI solutions, including Copilot, you can uh you can generate documents and artifacts and all these different like sort of more like richer things.
Showing 61–80 of 197 · page 4 of 10
← Previous
Next →