2026-06-28
The Working Physician's Personal AI Stack
A physician-executive’s take on the minimum viable AI stack that survives a real clinic week, and on the harder skill of knowing when not to use AI at all.
Read article →2026-06-28
Clinical AI Is a Governance Problem, Not a Chatbot Problem
A technically strong clinical AI model can still fail if no one defines ownership, workflow placement, escalation rules, and review cadence. In hospitals, the decisive question is not which model won the benchmark, but which team is accountable when the alert fires.
Read article →2026-06-26
AI Agents in Medicine Need Oversight, Not Blind Trust
AI agents may become the most useful junior member of the care team, but only if medicine redesigns supervision, escalation, and accountability around probabilistic systems. The hospitals that win here will treat agents like brilliant residents with bounded privileges, not autonomous clinicians.
Read article →2026-06-14
When the Algorithm Becomes a Second Opinion, Who Owns the Chart?
A physician-executive lens on why hospital AI should be governed like a clinical workflow, not a product category. The real risk is not that AI is too smart, but that hospitals adopt it faster than they can measure, supervise, and correct it.
Read article →2026-06-13
AI Agents in Medicine Need Oversight, Not Worship
AI agents can extend clinical and operational capacity, but only if medicine treats them like brilliant residents with limits, not autonomous replacements. The real challenge is organizational: building oversight, governance, and escalation pathways that match probabilistic systems.
Read article →2026-06-13
Future Proofing Medical AI Without Freezing It in Place
Medical AI should be stable enough to be safe and flexible enough to survive model drift, new evidence, and changing workflows. The right answer is not rigidity for its own sake, but a governed architecture that separates clinical intent, data plumbing, model logic, and deployment rules.
Read article →2026-06-11
What autonomous robotic surgery still gets wrong about embodied AI
I used to think the only real question in autonomous robotic surgery was when the robot would be “good enough.” Then I started looking at embodied AI outside the operating room, and the harder question became whether surgical autonomy can be certified, measured, and safely bounded at all.
Read article →2026-06-05
The hidden human cost of AI data labor
AI systems do not emerge from abstraction alone. They are built on invisible labor, and in medicine I think that fact changes how we should govern them, buy them, and trust them.
Read article →2026-06-01
The Hidden Labor Behind AI: Why Data Work Is the Real Ethics Test
AI systems do not emerge from abstraction, they are built on human labor that is often invisible, underpaid, and medically consequential. A physician-executive lens shows why hospitals and vendors must treat data labor supply chains as a governance issue, not a back-office detail.
Read article →2026-05-28
Why ambient AI scribes fail some clinicians, and what safe documentation automation actually looks like
Ambient AI scribes do not fail because documentation automation is impossible. They fail when hospitals treat them like a transcription purchase instead of a clinical workflow redesign with governance, fallback paths, and measurable supervision.
Read article →2026-05-22
Claude Headaches and the Human Brain at Maximum Bandwidth in Healthcare AI
The emerging problem is not that clinicians lack AI, but that AI now lets them run too many high-value workstreams at once. In healthcare, that creates a new kind of cognitive overload, with implications for safety, attention, and decision quality.
Read article →2026-04-21
Why Medical Tech UIs Feel Unintuitive, and Why Dog-Fooding Still Matters
Medical technology often feels clumsy because it is built around regulatory survival, fragmented legacy workflows, and limited clinical input rather than daily use. The fix is not prettier screens alone, but physician-led design, real-world testing, and vendor teams that use their own products in the environments where care actually happens.
Read article →2026-04-20
Cognitive Apprenticeship, Medical Residencies, and the Future of AI Model Alignment
Cognitive apprenticeship is not a nostalgic education theory, it is a practical blueprint for training both residents and AI systems to act safely under supervision. In healthcare, the same scaffolding that turns interns into clinicians may be the most useful model for aligning AI with physician judgment, workflow, and patient risk.
Read article →2026-04-01
What Hospital Systems Actually Need to See Before They Trust an AI Diagnostic Tool
Hospital leaders should not ask whether an AI diagnostic tool is impressive; they should ask whether it is clinically validated, operationally safe, and governable inside real workflows. Trust comes from evidence on performance, bias, monitoring, integration, and accountability—not marketing claims.
Read article →