The Ledger, every show

“We would see, like, 70% hallucination rates, like, 3.5 at the time”

“There was not a single like speech to speech model in existence at the time. And there was no research that was far along anything like that.”

“Now they have, like, a separate voice agent, like, platform that they've started to release, but it's a bit more abstracted. Whereas we're much more like low level, we expose all the raw pieces. So developers can like build like entire products on top of it.”

“The only models that could run at real time speed, like we're talking like under two seconds of, you know, to response time were like GPT, 3.5. Or like really shitty text-to-speech models.”

“And I'm sure Apple and Google, or sorry, Google and Amazon are gonna work on some very large enterprise facing versions of what we have.”

“Within six months, we were at about, like, four million in, in revenue annually recurring. Something, like, very small, small section of it was enterprise, now that's larger, but most of it was just, like, self-serve organic. We did zero ads. We did zero, like…”

“In like January, 20, 24, we had like 32,000 calls on the platform. If I even zoom back further than that, and that's all just sales roleplay training, basically.”

“Then we started to get some other folks like trickling in and we kind of hit like 52 K calls. We're looking at February. March was like 164,000 calls. We launched on product. So that was the first time, like, we ever launched publicly, and so immediately we ju…”

“And then we built this like button to join me in these really quickly. And it hadn't like 90% retention and it just like started growing and people just loved it.”