Skip to main content

Nicholas Nadeau

Aug 27, 2026

4:01
... and a, and a functioning Onnx version of them appearing in the app.
4:04
Yeah.
4:04
So it's, it's actually, it's not that difficult.
4:07
It's quite easy in a lot of ways.
4:08
A lot of the experts we work with are already, you know, published.
4:11
They have books, they have podcasts, they have substacks, blog posts, you name it, already have a lot of internet presence.
4:17
Now, this is a lot of the work that has been stolen and scraped by big AI in a lot of ways.

26 MINS LATER

30:51
So how do you, um, warn against that, think about that, um, how, and, and measure it, and then sort of mitigate it as a, you know, 'cause all AI systems have to deal with this at some point.
7:23
And that's a tall bar, and that's the one we've met, and I'll let you talk about how we got there.
7:27
Yeah, no, and, and it's really fun 'cause when you take a step back and really think about, you know, what does it take to build models and all that, we're probably one of the few companies that don't really have a data bottleneck.
7:36
We're working and building this trust with these experts, bringing in their public data, bringing in their dark data, and really creating a human-in-the-loop flywheel effect for, you know, self-improvement with these experts as part of it, the AI, not as a separate, you know, pitch it over fence or scrape from them and, you know, you create these alignment gaps.
7:54
You create these actual, you know, fidelity gaps in their expertise, how it's presented, and how it, you know, behaves to their intent.
8:02
And so there's really three layers of data that we'll bring in and, you know, one of the few companies that probably ask permission to use the data of these experts, you know, unlike big AI in a lot of ways.
8:12
And so we'll bring in, you know, when we start working with experts, we'll bring in your public data, all the books, your podcasts, your Substack, your blog posts, and a lot of the, these top experts in the world, especially in health and wellness, they have big public presences.
8:23
Most of our experts, you know, millions of followers, YouTube followers, Substack subscribers, books sold, you name it.

24 MINS LATER

32:46
Mm.
11:23
What is the core differences, you know, and you've touched on some of them, but can we expand on those differences?
11:31
Yeah, so we'll get into the evaluations, but I'll start with just high level, these massive, and when I say massive, they take an entire data center to run these giant models.
11:41
They are trained on the scraped, stolen, collected, everything of the internet.
11:47
So think about taking a vacuum to the entire internet.
11:50
Everything from the great things you find in Wikipedia to the not so great things you might find in Reddit.
11:55
All of that is part of the same data set vacuumed up, brought in, and trained on.
12:01
And so it's the statistical average of the internet, which experts are not an average person.

12 MINS LATER

23:43
How did you set that such that Onyx doesn't do that?

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.