The Ride Home Fund Profiles Onix: An Hour on Trust, Dark Data, and Ownership

The Ride Home Fund, one of Onix’s backers, featured Onix in a Portfolio Profile on the Tech Brew Ride Home Podcast. Co-founders David Bennahum and Nicholas Nadeau sat down for a full hour, which is long enough to get past the pitch and into how Onix actually works.

What the hour covers

The episode opens with a concrete case: what it means for a sports medicine physician like Dr. Jordan Metzl to train his own model, and how a person summons it. From there the conversation works through the parts of Onix that a shorter interview usually skips.

• Fidelity and trust. Why an expert’s onix has to sound like the expert, reason like the expert, and stay inside what the expert has approved. Generic AI averages the internet; an onix is scored on how close it stays to one person’s judgment.
• The three data layers. Public work, dark data, and implicit knowledge. The public layer is what a frontier model can scrape. The dark data, the unpublished notes and case material, and the implicit knowledge, the judgment an expert applies without writing it down, are what make the model worth training.
• Avatars of expertise, not digital twins. Onix builds an authorized model of how a person thinks about their domain, not a clone of the person.
• The business model. Subscriptions, not data. Experts set their own price and keep 70%.
• Privacy and on-device models. Conversations are encrypted on the person’s device. Onix cannot read them, train on them, or hand them over, because Onix does not have them.
• Ownership and the expert deal. No cost to join, no cost to leave, non-exclusive, and the model and data are deleted if an expert walks away.

The back half is the founding story: why David and Nicholas started with health and wellness, why high-stakes domains reward fidelity over breadth, and why they see the AI disruption as a parallel to the internet revolution, including where the balance between centralization and decentralization lands.

Why this format matters

Most coverage of Personal Intelligence® has to compress the argument into a paragraph. An hour lets the two of them show their work: where the training data comes from, what the expert controls, and what a person actually gets when they open the app. If you want the long answer to “how is this different from asking ChatGPT,” this is the episode.

Watch the full profile on YouTube, and see the rest of Onix’s backers on the Community page. To meet the experts building on Onix, visit onix.life.