HippocrAI is an open-source library of AI agents for medtech, built and maintained under physician oversight. This Substack is where the long-form writing lives — essays explaining how each agent works, the prompting patterns that make them safe, and the discipline that keeps them releasable.
The agents themselves are at github.com/hippocrai. They are free, Apache-licensed, and built so that a medtech founder, an academic researcher, or a curious engineer can clone any of them and have working software within ten minutes.
This publication exists because the most useful AI tools in medicine are not going to come from AI labs that don’t understand medicine, or from medical institutions that don’t move fast. They will come from physicians who have learned to use AI as a force multiplier and who hold themselves to the discipline the profession demands. HippocrAI is what that work looks like when it’s done in public.
What you’ll find here
Long-form essays on building AI tools for medtech. Each new agent ships with a companion essay — typically 1,500–2,500 words — walking through the architecture, the failure modes the design defends against, and the choices that look small but actually matter. These are written for physicians and other domain experts who want to understand how this kind of tooling works, not just use it.
Prompting patterns for domain experts. Reusable techniques for getting AI tools to produce safe, calibrated output in regulated contexts — medical, legal, regulatory, scientific. The patterns generalize beyond medtech, but they are written from inside it.
Notes on the surrounding discipline. IP firewalls. FDA-compliant framing. Patient-safety language. Pre-release review protocols. The unglamorous infrastructure that turns clever prompts into releasable tools.
No marketing emails. No pitch decks. No course launches. This is a publication, not a funnel.
What this is built around
Open by default. The agents are free. The Substack subscription is free. The thesis is that the moat in medical AI tooling lies not in the code, but in the discipline applied to it. Anyone with the discipline can build their own.
Physician-led. Every artifact published here is filtered through the constraints medical training instills — patient safety language, regulatory awareness, IP firewall, verification expectations. These are not afterthoughts. They shape what gets built and what gets released.
Medtech-specific. General-purpose AI tools fail at medtech work because they don’t know what good looks like in this domain. HippocrAI’s tools are tuned for the unglamorous middle of medical device development — grant writing, regulatory drafting, literature review, IP support, founder operations.
Built under the oath. First, do no harm is the operating principle, not a brand tagline. Tools that could leak provisional patent content, frame the agent as a medical device, or enable misuse in clinical decision-making are not released.
Who’s behind this
HippocrAI is written and maintained by Joseph L. Hayhurst, MD — a physician-founder with prior training in general surgery, currently focused on medical device development. The work here is independent of any clinical institution or commercial venture and contains no proprietary content from private projects.
I write because the patterns I use to build these tools are usable by other physicians, and because more physicians being able to build their own tooling is good for medicine. I publish openly because a thesis that lives only in private repositories cannot be challenged or improved by anyone but its author.
How to engage
Subscribe to receive new essays as they’re published. Free.
Read the code at github.com/hippocrai. Use it, fork it, build on it, or rebuild it differently.
Reach out with feedback, questions, or examples of agents you’ve built using these patterns. Email me at joseph.l.hayhurst@gmail.com or open an issue on any HippocrAI repository.
The agents are free. The discipline is the moat.
— Joseph L. Hayhurst, MD
