Onix's 'Personal Intelligence': Expert Curation or Wellness Vaporware?
The Montréal startup claims a privacy-first alternative to frontier AI, but its 'expert' library relies on content creators rather than clinical peer review.
In the current gold rush of generative AI, the promise of 'personalized health' is often a thin veil for generic large language models (LLMs). Onix, a Montréal-based startup founded in 2025 by David Bennahum and Nicholas Nadeau, claims to be the antithesis of this model. As BetaKit first reported, Onix is building a library of "onixes"—AI versions of health and wellness practitioners trained on specific datasets rather than the open web.
On paper, the mechanism is straightforward: Onix partners with wellness content creators who provide public appearances, such as podcast recordings, and unpublished works to craft their digital twins. The library spans fields from longevity and addiction recovery to holistic nutrition, featuring practitioners such as psychiatrist David Rabin and family nurse practitioner Ashley Shrader.
From a technical standpoint, the company is attempting to dodge the privacy pitfalls of frontier models. Co-founder David Bennahum told BetaKit that the company utilizes smaller models run locally on users' devices, eliminating the need for conversation logs to be sent to a data center. Bennahum claims this approach reduces inference costs to "one two hundredths" of a frontier model.
However, for those of us wary of the "wellness" label, the critical question is the nature of the source data. Onix is not positioning itself as a clinical tool; Nicholas Nadeau explicitly stated to BetaKit that the app does not provide medical advice and employs guardrails to prevent the AI from straying into that territory. Instead, the platform operates as a marketplace for "personal intelligence," where practitioners set subscription fees and receive 70 percent of the revenue.
While the privacy architecture is a welcome departure from the data-harvesting norms of Big Tech, the "expert" designation remains loosely defined. The training data is curated by the practitioners themselves—often drawing from podcasts—rather than being grounded in peer-reviewed clinical trials. This suggests that the value of an "onix" is tied more to the practitioner's personal brand than to a standardized medical gold standard.
Despite these caveats, the market appetite is evident. Onix recently closed a $5 million USD ($6.9 million CAD) pre-seed convertible note round led by Los Angeles-based AlphaEdison, with participation from Garage Capital, the Ride Home Fund, Jeremy Zimmer, and Jean-Sébastien Cournoyer. As Onix launches its paid model for iOS users, it is betting that consumers will prioritize the perceived authority of a named expert and the security of local data over the raw power of a frontier LLM.