Sigil Wen has made a significant move in the AI space by launching Underdog, a personal assistant that emphasizes user privacy by processing data directly on users’ devices, specifically Macs and Windows PCs. This approach stands in stark contrast to AI assistants like Meta’s Muse, which depend on cloud processing, making Underdog a notable contender in the market. Currently, the app is in an invite-only beta phase, with plans to expand to Linux, iPhone, and Android in the future.
A self-taught coder, Wen relocated to Silicon Valley at 17, immersing himself in an environment rich with innovation. He lived in an AI hacker house with famed AI researcher Andrej Karpathy and other people who would become the biggest names in AI, like Perplexity founder Aravind Srinivas and OpenAI researcher Noam Brown. During this formative time, Wen engaged with early versions of significant AI tools, including a chatbot shared by Anthropic co-founder Ben Mann that would become Claude, an image generator from David Holz that would become Midjourney, and what would become OpenAI’s GPT-3 and the image generator Stable Diffusion. For fun, he also got GPT-2 running on his Apple Watch. “It was a magical time,” Wen recalled.
Now a Thiel Fellow – the program from investor Peter Thiel that invites young founders to pursue projects instead of college – Wen built Husky, an inference engine designed to run AI models fast. Unlike other on-device engines, Wen says, Husky moves less data between the computer’s main chip and its graphics chip. Underdog also encrypts the keys to email and other accounts that users authorize it to access. Wen developed Husky, an inference engine designed for speed, which reduces the amount of data transferred between the computer’s main chip and its graphics chip. This efficiency is essential for maintaining privacy while delivering functionality.
Currently, Underdog uses a 27-billion parameter reasoning model fine-tuned from Qwen3.8 27B. While it employs smaller models compared to those hosted in data centers, Wen asserts that it can still effectively handle tasks like shopping research and solving math homework questions. “You don’t need to sacrifice your privacy for the capability because they’re just as capable,” he stated, emphasizing that small on-device models will continue to grow more capable over time.
The business model for Underdog is equally fascinating. Initially, the app will be free and will never include ads, distinguishing it from many competitors who rely on data mining for revenue. Since the AI operates directly on user devices, Wen avoids the hefty costs associated with cloud services, allowing him to forgo subscription fees. Instead, Underdog plans to take a tiny percentage from payment transactions made through Stripe’s secure payment rails, similar to an interchange fee. This approach aligns Wen’s interests with those of the users, as the AI assistant won’t need to exploit personal data for profit.
Wen’s vision for Underdog reflects a personal commitment to creating a product he would be proud to share with future generations. “I honestly want to build Underdog for myself. I’m building a product that I would be proud for my future children to use,” he shared. Supported by notable investors like Patrick Collison, co-founder of Stripe, and firms such as Andreessen Horowitz and Khosla Ventures, Wen is positioning Underdog as a reliable alternative to existing AI assistants that often sacrifice user privacy for functionality.
Underdog’s business model, which avoids ads and subscription fees by taking a small percentage from payment transactions, directly reflects its commitment to user privacy while maintaining functionality. This approach not only sets it apart from competitors that rely on data mining but also aligns Wen’s interests with those of users, emphasizing a growing trend toward privacy-centric tech solutions.
