Doctors plus AI, not big tech plus your data

Why "curbside"

A curbside consult is the oldest good tool in medicine: you catch a colleague in the hallway, describe the case in thirty seconds, and get a straight answer from someone who has seen it before. No consult order, no wait, no bill. It works because it is fast, specific, and trusted.

That is the interaction we are rebuilding — fast, specific, available at 2 a.m., and, crucially, safe to be specific with, because the conversation does not leave your control.

The problem we are actually solving

Clinical AI got good by consuming clinical work: notes, conversations, orders, outcomes, and the questions physicians ask when they are unsure. Once that data is de-identified, HIPAA stops governing it, and whoever holds it can build what they like with it — including systems designed to do the clinician's job without the clinician.

That is not hypothetical. A federal bill would let software qualify as a prescribing practitioner. A model bill circulating in state legislatures would issue AI its own license, up to fully autonomous diagnosis and treatment. Utah has already waived its rules so that one company's AI can process prescription renewals. The record, with primary sources, is on Why it matters.

Meanwhile clinicians face a bad choice: use the AI tools and feed the pipeline, or refuse them and work slower than the colleagues who don't. We think that is a false choice, and that the third option is not abstinence but ownership — a pipeline that runs back to the people feeding it. That is a product and governance problem before it is a policy problem.

What we believe

1. AI does not have a license to practice medicine. A doctor equipped with strong AI is safer and more effective than AI alone — and unlike AI alone, that doctor can be held accountable.
2. Clinical data belongs to the clinician and the patient, not to whichever software happened to be in the room.
3. De-identification is not consent. "It's legal" is a floor, not a standard. Clinical data is going to build the next generation of clinical AI — the question is whether anyone asked, and whether the people who created it have a say in what gets made. So we ask, and the answer is reversible.
4. The access problem is real. Shortages, rural coverage, and after-hours gaps are genuine, which is why the autonomy pitch lands. We think making one clinician dramatically more capable answers it better than removing the clinician.
5. Trust is a contract term, not a value. If it isn't in the agreement, it isn't a commitment. Hold us to that.

What we're building

It is easy to state principles and quietly leave room to change your mind, so here is the list, specifically:

• A secure frontier model you can be specific with — chat, document read and write, and an ambient scribe, under a BAA with your practice
• Data that lives in your environment, on retention you set, with full export and verifiable deletion
• A collective pool that improves these tools, that you choose to join and can leave
• Safety, quality, and bias monitoring, so "trust it" is something we measure rather than assert
• Practical material on where AI actually belongs in a clinic day — what to hand it, what to keep, how to check it
• A structure where the physicians using the tools direct what gets built next
• A path from getting your evenings back to running a practice that delivers care at scale

And one thing that stays off the list, because it is the point of everything above: every product we build keeps a licensed clinician in the loop. No autonomous diagnosis, treatment, or prescribing — no product, no roadmap item, no pilot — and no application for an AI practitioner license under H.R. 238-style or model-bill frameworks.

Who this is for

Independent practices, small groups, and clinicians inside larger systems who want a tool of their own — one whose terms they negotiated and can read. Primary care, specialty clinics, and anyone whose day is half documentation and payer correspondence.

Where we are

Early. We are onboarding a limited number of practices and building around what they actually do all day, rather than shipping a general-purpose assistant and hoping medicine adapts to it. If that is something you want to be early to, tell us what you would point it at first.