Baymax

Company: Baymax (self-reported, unverified)

What it appears to be: A local-first health intelligence system for people with chronic conditions, built as a prototype for the OpenAI 2026 hackathon. It integrates clinical data, wearables, and patient narratives into a structured longitudinal memory system that uses GPT-5.6 to surface evidence-linked insights and trigger proactive check-ins.

What changed: The author states they built this as an individual over one intensive weekend, using Codex as their engineering collaborator. The project evolved from concept to a working native mobile application with backend infrastructure supporting FHIR, wearable data, and semantic memory.

Single most important open question: Is there evidence of traction, revenue, or customer adoption beyond the author's prototype?

Confidence level: Very low — this is entirely self-reported, unverified information with no third-party corroboration. The description contains no evidence of commercial activity, customers, revenue, or market validation.

The description states that Baymax is a "local-first, longitudinal health-intelligence system for people living with chronic conditions."

It connects:

• Clinical data (conditions, medications, encounters, lab results, documents)
• Wearable data (sleep, activity, HRV, workouts, measurements)
• Patient narratives (qualitative reports in original words)
• Deterministic personal baselines

The system uses GPT-5.6 through the Vercel AI SDK to retrieve and synthesize information from these sources, but does not directly calculate trends or query databases. Instead, it calls typed tools for specific data retrieval.

It is described as a functional Expo application for iOS and Android connected to a local Hono coordinator that manages authentication, data contracts, model routing, clinical connections, longitudinal memory, and proactive jobs.

Evidence: The author's own description.

Inference: This appears to be a prototype built for a hackathon with no commercial deployment or user base.

The author states Baymax is:

• A "continuously maintained, patient-owned health narrative"
• Not a dashboard, tracker, or chatbot
• Designed to preserve patient language and avoid collapsing different types of information into one model context
• Focused on understanding what changes between clinical visits rather than providing diagnosis or treatment

It positions itself as distinct from typical health apps by:

• Separating authoritative clinical truth from semantic memory
• Using deterministic personal baselines instead of population averages
• Preserving patient narratives as authoritative sources
• Making observations traceable to evidence
• Exposing model uncertainty and data boundaries

Evidence: The author's own description.

Inference: This is a positioning statement for a health intelligence tool, not a claim of market traction or adoption.

The description states that Baymax targets:

• People living with chronic conditions
• Patients who see clinicians four times a year but whose conditions do not disappear during the other 361 days
• Individuals who experience symptoms, sleep deterioration, medication changes, and lifestyle variations between clinical visits

It also mentions "patient-caregiver households" as a future validation target.

Evidence: The author's own description.

Inference: The ICP is individuals with chronic conditions seeking better longitudinal health management, but there is no evidence of actual customers or user testing beyond the prototype.

The description does not contain any information about:

• Revenue streams
• Pricing models
• Monetization strategy
• Customer acquisition costs
• Unit economics

It only describes a local-first architecture that can be deployed on patient-controlled hardware or hosted environments, but says nothing about how this would generate value for users or customers.

Evidence: Not evidenced.

Inference: No business model or pricing evidence is provided in the self-reported description.

The system is built using:

• Expo (iOS/Android mobile app)
• Hono coordinator
• Vercel AI SDK with GPT-5.6 Terra
• FHIR R4, SMART-on-FHIR, HealthKit, Health Connect
• PostgreSQL for persistence
• Supermemory for semantic recall
• Docker containers
• TypeScript

It supports:

• Deterministic personal baseline calculations
• Evidence-linked qualitative memory
• Proactive evaluation and outreach controls
• Consent enforcement
• Audit records
• Local-first operation with multiple runtime profiles (fixture-only, local-product, connected-provider)

The system is described as supporting both patient-controlled and hosted deployment.

Evidence: The author's own description.

Inference: Technical architecture is detailed but lacks evidence of production use or scalability beyond prototype.

The description states:

• This was built as a hackathon project
• It includes a working native product experience
• It has a real GPT-5.6 tool-calling agent
• It contains deterministic personal-baseline calculations
• It supports synthetic but transparent demo data
• It demonstrates one complete loop: records + wearables + lived experience → baseline change → constrained decision → consent-checked outreach

However, there is no evidence of:

• Revenue or monetization
• Customer adoption or usage metrics
• Market traction
• Product-market fit validation
• Any real-world deployment beyond the prototype

Evidence: The author's own description.

Inference: This is a prototype with no demonstrated traction or maturity in a commercial context.

The description states that most health applications are dashboards, trackers, or chatbots. Baymax positions itself as different by:

• Not forcing experiences into scores
• Preserving patient language
• Comparing individuals against their own baselines rather than population averages
• Separating clinical truth from semantic memory
• Making observations traceable to evidence

It does not name specific competitors but implies a gap in the market for more nuanced, longitudinal health intelligence systems.

Evidence: The author's own description.

Inference: No competitive analysis or market positioning beyond self-stated differentiation is provided.

Key risks and red flags based on the description:

• Entirely unverified claims about functionality and performance
• No evidence of revenue, customers, or adoption
• Prototype-only development with no commercial deployment
• Heavy reliance on GPT-5.6 without clear evidence of how it scales or is monetized
• Patient-owned architecture may be difficult to commercialize at scale
• Local-first approach could limit user accessibility and data portability
• No mention of regulatory compliance, security, or privacy frameworks beyond consent enforcement

Evidence: The author's own description.

Inference: These are potential risks based on the lack of traction, commercialization, and real-world validation.

1. What is the actual user base or pilot group for this prototype?
2. How does the system handle regulatory compliance (e.g., HIPAA, GDPR)?
3. What are the technical limitations of the current architecture that would prevent scaling to real users?
4. Has there been any external validation or feedback from clinicians or patients?
5. What is the long-term plan for monetization and product development beyond the prototype?
6. How does the system ensure data integrity and privacy across multiple deployment environments?
7. What are the key assumptions about user behavior that underpin this design?
8. Are there any known technical constraints around wearable data integration or model inference?

Evidence: Not evidenced — these are questions to probe the unverified claims.