What I’m Doing Now

Last updated: Wednesday, July 1, 2026

This is a now page. It changes when I do.

Quick Snapshot

Since the last update, a lot moved from idea to artifact. I delivered the AcademyHealth talk in Seattle, gave a virtual talk to the NIH Citizen Science Working Group, recorded two health AI podcast interviews, shipped several public patient-directed AI projects, and am now preparing to fly to Portugal for the closing keynote at ACM Interactive Health '26.

The through-line is patient agency. I am building and explaining tools that help patients and caregivers use AI to understand health data, challenge institutional blind spots, and act with more confidence.

The short version: critical AI health literacy is no longer just a framework for me. It is becoming software, websites, scorecards, talks, and community practice.

Heading To Portugal

On July 8, 2026, I am giving the closing keynote at ACM Interactive Health '26 in Porto, Portugal.

The title is:

Same Technology, Different Master: AI as Liberation Technology

This is the full-length version of a talk arc that started with AcademyHealth in Seattle on May 30 and continued with the NIH Citizen Science Working Group on June 16. The Porto audience is especially interesting because it comes from human-computer interaction, participatory design, and health technology. That is exactly where the patient-agency question needs to be.

For years, health technology has often meant software built by institutions, researchers, vendors, or clinicians and pointed at patients. Even participatory design can leave patients in the role of participant rather than researcher, subject rather than builder.

My argument is different: patients are already directing AI to build the tools their institutions never would. The question is whether health technology will recognize that shift, support it, and help make it safer and more equitable.

What Shipped Since May 4

OpenKP

OpenKP is my patient-directed MCP server for Kaiser Permanente Northern California's patient portal. It runs locally, uses my own credentials, and lets Claude work with my own record under my control. The project now has a public GitHub repo, an MIT license, CI, and a working set of tools across appointments, labs, messages, medications, problems, allergies, demographics, visit notes, and after-visit summaries.

This is the concrete artifact behind much of my recent speaking. It is one thing to argue that patients need agency. It is another thing to show a patient-owned bridge to a health record that actually works.

My Heart Data

I built myheartdata.org, a public dashboard of my hypertrophic cardiomyopathy echo history. It was built in one day, using patient-directed AI collaboration. I pulled together 19 echocardiogram studies from four sources spanning 2007-2026, normalized them, redacted the public data, and published a longitudinal view a cardiologist can actually read.

The dashboard tracks key HCM markers over time, shows missing data as data, includes point-level auditability, and makes visible the kinds of longitudinal patterns that patient portals do not show.

This is CAIHL in practice: the patient defines the question, selects the data, controls the pipeline, and decides what can safely be public.

HugoScore

I launched hugoscore.org, a public prototype for evaluating health AI tools by whether they increase or decrease patient agency.

Most health AI evaluation looks at accuracy, efficiency, adoption, compliance, or clinical workflow. HugoScore asks a different first question: who does this tool serve?

The current prototype includes a searchable directory, public submission flow, draft CAIHL profiles, evidence links, confidence labels, and a human-review-first publishing model. It covers health AI that affects patients whether or not patients directly see the tool: ambient scribes, patient health copilots, record-access tools, navigation tools, triage systems, and more.

The key principle is simple: automation can help research and draft evaluations, but human review is required before making public claims about safety, privacy, equity, or patient impact.

CAIHL Skill

I packaged the Critical AI Health Literacy framework as a reusable Claude Skill and launched it at caihl.org.

The skill applies CAIHL as a way of seeing. It can be used to analyze news, evaluate products, critique papers, or draft content about institutional AI, patient-directed AI, algorithmic resistance, ambient scribes, prior authorization AI, clinical decision support, and health AI equity.

The repository is now public. The skill is usable as a zip upload in Claude or as a folder-based skill for local agents.

This matters because the framework should not live only in papers and talks. It should travel into the tools people actually use to think.

Endothelial.org

I built endothelial.org, a plain-language knowledge base about endothelial dysfunction, COVID, long COVID, and ME/CFS for Dave deBronkart, also known as e-Patient Dave.

Dave handed me a pile of papers and needed something usable while dealing with post-COVID cognitive strain. The result is a gentle, accessible resource with a primer, glossary, FAQ, reading guide, themes, text-size controls, display themes, collapsible sections, audio overviews, and a public companion NotebookLM.

This was built for one person first. That is the point. Patient-directed AI is often most powerful when it starts with a real person, a real need, and a design rule like: reduce cognitive load, never add to it.

OpenEP & Hugotronic

OpenEP, live at hugotronic.com, remains my patient-built dashboard for ICD interrogation data.

It ingests every interrogation report I have collected since 2007 and renders a single longitudinal view of cardiac device telemetry across three generators and the same long-lived leads.

Institutional tools do not give me that view. OpenEP does.

Together, OpenEP and My Heart Data show the same pattern from two angles: patients can build clinically useful views of their own longitudinal data when they have access, tools, and enough critical literacy to know what questions matter.

Talks, Podcasts, And Writing

I delivered my AcademyHealth Annual Research Meeting talk in Seattle on May 30. The core distinction was institutional AI versus patient-directed AI, with the downskilling of doctors and upskilling of patients as the tension underneath.

On June 16, I gave a virtual talk to the NIH Citizen Science Working Group on patient-directed AI, critical AI health literacy, and turning fear into agency. That talk sharpened the NIH question I keep returning to: in an industry-driven AI landscape, what is the public role, and where is the patient voice?

I also recorded a Practical AI in Healthcare podcast episode with Dr. Steven Labkoff and Leon Rozenblit. The episode is live as "Hugo Campos: Patient-Directed AI." I recorded another interview with Pulse Pod in Australia, focused on patients using AI and the practical reality behind the public examples.

The co-authored chapter with Larry Chu, "The Third View: What Patients Seek When They Turn to AI," is with the editor. A separate coauthored manuscript is also under journal review.

The public story is catching up with the lived one.

AI Patients

AI Patients is still the community layer of this work: patients and caregivers learning to use AI as a thinking partner for health, not to replace doctors, but to ask better questions, spot patterns, translate complexity, and prepare for care.

The first meeting on March 13 showed the need clearly. Patients and caregivers from across the U.S. and Portugal joined from lived experience with genetic heart disease, diabetes, cancer, lupus, cystic fibrosis, long COVID, chronic Lyme, and complex family caregiving.

The community cadence has been quiet while I have been shipping talks and tools. I am thinking carefully about how to restart it in a way that is easier to sustain: practical demos, shared workflows, examples people can actually reuse, and a publishing home that keeps the conversation visible.

What I Am Working On Now

Right now, Porto is the main focus. The message is not that patients should be left alone with AI. The message is that patients are already using AI because the existing system leaves too much unexplained, inaccessible, fragmented, or delayed. The ethical task is not to pretend that is not happening. It is to help patients do it critically, safely, collectively, and with agency.

What I Am Looking For

I am especially interested in:

  • Speaking and workshop opportunities on patient-directed AI, critical AI health literacy, and local-first health data tools

  • Research collaborations on how patients and caregivers actually use AI with their own records

  • Funders and partners who understand that patient AI should not be limited to institutional portals

  • Technical collaborators working on FHIR, MCP, local agents, privacy-preserving health tools, and patient-owned infrastructure

  • Patient and caregiver communities already using AI for health who want to compare notes

  • Health AI teams willing to have their tools examined through a patient-agency lens

Tools I Am Using

  • Claude Code CLI for serious build work.

  • Claude Cowork for project planning, document drafting, reading records, preparing talks, and working with local files on my Mac.

  • Claude for Chrome for browser-side tasks where the data is already on the page.

  • Gemini, NotebookLM, and ChatGPT for cross-checking, research, and triangulation.

  • Wispr Flow for voice input in English and Portuguese. Most of what I write starts as speech.

Ideas I Am Working With

The gap between data access and data comprehension. Federal rules made the data accessible. They did not make it understandable. AI can be the bridge.

Patient agency as an evaluation criterion. Accuracy and efficiency are not enough. Health AI should be judged by whether it expands or constrains the patient's ability to understand, question, decide, and act.

Local-first AI. The data stays on your machine. The credentials are yours. The agent answers to you.

Community as infrastructure. The most interesting health AI work right now is happening between patients, not inside health systems.

The speed of desperation. As my friend and fellow advocate Sue Sheridan puts it: "Doctors adopt AI at the speed of trust. Patients adopt it at the speed of desperation."

Patients are not waiting for institutional permission. They are already here.

If any of this resonates, I would love to hear from you.

Let's Work Together