Oct 28, 2026 · 44 min · 13 segments
William Collins and Eyvonne Sharp discuss how smaller models change AI application economics, when hybrid routing and caching pay off, and why multi-call agents can still produce large cloud bills…
Eyvonne SharpHostWe're, you know, AI wasn't born yesterday, although for some of us, it might feel like it.
There's starting to be some, you know, like real operating models that have been assembled for how one might operationalize models.
And with that, I think the, to kind of like kick this off the There's this thing that sort of happens where it's kind of like in the days of when we were provisioning VMs in the data center where you just want the biggest and best that you can possibly get that you can possibly request for whatever you're doing.
When really, as we've started to mature here, that smaller model pricing is getting very interesting.
So I guess in the first, as we start thinking through this, kind of the chain of thought through when, why, how, and all those things, You might begin to start thinking like what becomes practical when inference gets cheaper like this from the frontier model providers.
I'd start with the stuff that maybe sometimes like we leave on the table because it didn't justify that expensive model call.
We're, you know, AI wasn't born yesterday, although for some of us, it might feel like it.
There's starting to be some, you know, like real operating models that have been assembled for how one might operationalize models.
And with that, I think the, to kind of like kick this off the There's this thing that sort of happens where it's kind of like in the days of when we were provisioning VMs in the data center where you just want the biggest and best that you can possibly get that you can possibly request for whatever you're doing.
When really, as we've started to mature here, that smaller model pricing is getting very interesting.
So I guess in the first, as we start thinking through this, kind of the chain of thought through when, why, how, and all those things, You might begin to start thinking like what becomes practical when inference gets cheaper like this from the frontier model providers.
I'd start with the stuff that maybe sometimes like we leave on the table because it didn't justify that expensive model call.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.