Frontier AI Models Now Fit on Your MacBook | Sigil Wen artwork

Frontier AI Models Now Fit on Your MacBook | Sigil Wen

MTS

October 6, 2026

Sigil Wen, founder of Conway Research and Thiel Fellow, discusses launching Underdog—a private, local AI running directly on personal devices—and how autonomous agents will transform internet commerce by rendering traditional advertising obsolete. Turn ideas into software people love.

Speakers MTS, Sigil Wen

TopicsNews

MTS (0:00)

All right, we are back, we're live with Sigil Wen, who is a Thiel fellow. He is the founder of Conway Research, and they just announced their newest product, Underdog, which is really cool. I got an early look at it, because I was thinking of places to do work on my laptop in SF a few weeks ago. And I was like, oh, I really love the one hotel, West Hollywood, when I stayed there. Oh, there's one in SF, maybe I will hang out there. So I go to the hotel, and sitting in the lobby was Sigil, who was there working on Underdog. And so I downloaded it on my laptop, and I got to experiment with it a little bit, and it was really cool, actually.

Sigil Wen (0:41)

Thank you so much, Thiel. Yeah, I was working on Underdog from the one hotel, so I was like, what, Thiel, you're here.

MTS (0:47)

Yeah.

Sigil Wen (0:49)

So yeah, Underdog is your private personal AI that runs on devices that you already own. And I really believe that there should be no trade off between privacy and capability. And that used to be, you know, you used to have to trade off privacy and capability, but now with Underdog, you don't have to.

And what's really crazy, and I've really been tracking the Pareto frontier of AI models, and I've noticed this pattern where now the frontier models that could run on your devices and can fit on your MacBook, can be more powerful than the frontier models that OpenAI Anthropic released from more than six months ago. And I call that Underdog's law, where the frontier models from today, you will be able to bring to devices that run fully privately and locally to everyone.

MTS (1:49)

Yeah, it's a really cool vision. How did Conway Pivot from its previous product to Underdog? It was really interesting. I remember at the launch, it was about agents on the web. It was very prescient at the time. Nobody really had been thinking about this at all. I love the reference to Conway's Game of Life.

How did you go from A to B?

Sigil Wen (2:14)

Really, it is all the same. It will be very, very apparent.

Yeah, we were looking into how agents can access the world and can interface with the economy. I think that there is going to be a radical reformation of the economy as businesses and the economy shifts to serving agents as one of its newest users on the Internet, especially as humans trust AIs more to act on their behalf, to be able to purchase things for them.

It's super different. So I felt that security and privacy is really important, and it's something that I personally care about. We were working on agents that could run on long horizon harnesses. And we were also testing different frontier models. And it was just really scarily apparent just how good at cyber they were. And to the point where I think it was really clear to me that, oh my god, you know, Mithos class models are open source now. And I think that really led to me wanting to build Underdog for myself. And Underdog really was like, I want to bring my data onto my devices off of, you know, data centers, which are some of the biggest targets. If you don't trust, you know, like the real question is, if something that could run fully on your device, just as good exists, like why wouldn't you use it? Why do you need a trade off? Like I need to stream my voice. I need to stream my therapy sessions or coaching or, you know, my math ideas to some data center for AI training.

For instance, with meeting recording. But that for instance, is a functionality within Underdog where like, I don't want to do that. I don't think other people want to do that. And you don't need to do it because you can actually have model that is just as good, if not better, running faster because it doesn't have to do that round trip to the data center. And yeah, like I think that as humans interface and trust agents to transact on their behalf, the entire economy of the internet shifts, which is a lot of what I've talked about in my earlier web four post, where you no longer have to follow these like creditory business practices, like advertising, for instance, right? And like as the effort of doing certain tasks goes to zero, and on intelligence that you actually own, it changes the entire economy, where suddenly underdog can within a couple seconds, scan the price of a flight across every major website on the internet, can find the cheapest price, and then can generate a new virtual card to get it for you, thus saving you money and also saving you time. And agents that are truly aligned with you, you know, they don't have to like, advertising is such a boomer, like I see the business model and you don't have to do that. And it changes, right? Like agents are rational. They're just going to choose whatever is cheapest, whatever is the best, and they can do it effortlessly because they could scan the whole internet for it. And I think businesses are going to have to adapt to the fact that consumers are using agents like underdog to, you know, act on their behalf. And they are susceptible to advertising and all these dark existing practices that are mostly for humans as users of the internet.

3 more minutes of transcript below

Thousands of transcripts fetched by people building searchable podcast archives

Fetch the whole transcript

The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000793504734