Mantis

Market position

Mantis argues that Google, Apple and Meta each hold only a deep slice of the human record, see only their own products, and monetize through a single narrow business, leaving a holistic model of the individual structurally outside every incumbent's mandate. It compares its intended role to OAuth for identity, Plaid for banking and Stripe for payments — an intermediary that does not compete with either side of the exchange.

The company's stated differentiators are the physics-engine layer that grounds generated synthetic data in realistic anatomy and physiology — allowing datasets to be generated for edge cases with no public data — combined with an LLM layer for reconciling disparate, unstructured sources, and a consent architecture that delivers scoped, revocable predictions rather than raw data.

Technology

The system assembles a per-person record from continuously measured streams (device telemetry, wearable physiology, digital activity, clinical records) and applies an LLM-based layer to route, validate and synthesize heterogeneous inputs, then runs the data through a physics engine to produce high-fidelity, physically grounded renders and synthetic datasets used to train predictive models. The company's stated thesis is that raw human behavior is chaotic and unforecastable, but that correct coarse-grainings of it are not: stable aggregates (resting-heart-rate trends, sleep architecture, adherence rates, cognitive baseline drift) and stable routines (sleep location, commute, spending categories) are forecastable, while micro-chaos such as the next word typed remains a longer-term goal. Mantis cites published research as evidence for the achievable ceiling, including a 93% upper bound on the predictability of individual location (Song et al., Science, 2010), personality inference from ~300 Facebook Likes (Kosinski et al., PNAS, 2013 and 2015), and a finding that 81% of COVID-19 cases showed physiological warning signs at or before symptom onset, some nine days early, from a smartwatch alone (Mishra et al., Nature Biomedical Engineering, 2020).

Go-to-market

Mantis began with professional sports customers, including an NBA team, while planning to broaden to pharmaceutical labs and FDA trial researchers and, later, a general-public preventative-health product. The company said its March 2026 seed funding would go toward hiring, advertising, marketing and go-to-market functions. It was recruiting a founding engineer in New York and a part-time content creator/video editor as of the Y Combinator listing.

Builders and companies that need forecasts about individual people, initially in healthcare: care teams and providers, clinical researchers, pharmaceutical labs and researchers working on FDA trials. The company has also sold into professional sports, with an NBA team cited as one of its main clients, and says it eventually intends to release the platform to the general public for preventative healthcare.

Geography

Headquartered in New York City, New York, United States, with job listings for New York and remote (US) roles.

History

Mantis was founded in 2025 and is based in New York City. It participated in Y Combinator's Winter 2026 batch, with Gustaf Alstromer as primary partner, and had a team size of three as listed on its YC profile. In March 2026 it announced a $7.4 million seed round.

Risks & controversies

Mantis's model depends on aggregating sensitive personal physiological, behavioral and clinical data, an area subject to privacy expectations and regulatory constraints; the founder has stated that people's data should not be exploited and that access is scoped and revocable. The company also notes that ethical and regulatory constraints limit the availability of patient data for public datasets and AI training, which is the gap its synthetic-data approach targets.