Doug Fullington

AI Did Not Beat the Doctor. It Beat the Hour. open.substack.com/pub/dfullingto…

I ordered a workup last month that I thought I knew cold. Checked the current guidance first. The field had moved. The patient got a different workup than she would have gotten twenty minutes earlier.

@EMostaque Always Extended.

Do you use normal or extended thinking for GPT 5.5 Pro

Information attached to a real patient retains better than anything consumed in a webinar. The face across from you is a more powerful mnemonic than any flashcard.

The bottleneck in medical education was never information. It was retrieval at the moment of decision, with a patient in the room and four more waiting.

I dismissed EHR alerts for fifteen years because the signal-to-noise ratio made reading them irrational. That is not a physician problem. That is a design problem. The tools I actually use now are ones I asked for.

A 36 percent end-to-end task completion rate for the best AI agent on administrative workflows. The prior authorization piece is not ready to hand off. The documentation piece is. Know which layer you are working in.

The EHR is not being replaced. It is becoming infrastructure. The difference matters more than it sounds.

The EHR Is Being Hollowed Out open.substack.com/pub/dfullingto…

The EHR Is Being Hollowed Out open.substack.com/pub/dfullingto…

Fifteen years of medical training selected for confident recall under pressure. That's not what AI rewards.

Pro tip: Write your best content in batches, then schedule throughout the week. Your audience stays engaged, and you stay sane. Time to create 💪

Also try these power features: ✦ Press ⌘J (Mac) or Ctrl+J (Windows) to rewrite with AI ✦ Drag & drop media directly into the editor ✦ Use the top right buttons to share & organize drafts ✦ Check out the Calendar & Analytics pages

Hey, this is a sample draft, try editing it. Here's how Typefully helps you build an engaged audience: 📝 Write without distractions ⏰ Schedule for peak engagement 📊 Track what resonates 🚀 Enable auto-engagements 🔄 Cross-post everywhere

A patient watched me ask an AI to critique its own answer. 'You can make it check itself?' She left that visit with a skill she can use every time she searches her symptoms online.

Ninety percent of clinical questions I had were never answered. Not because I didn't care, but because the cost of looking them up exceeded the time I had.

@Dostoevskyquot Redemption

describe this book in ONE word

Stop Teaching. Start Learning. 90% of my clinical questions went unresearched for years. Not because I did not care. Because the cost of asking was too high. When that cost dropped to 30 seconds, everything changed. New article on why physician AI education has the pedagogy backwards: open.substack.com/pub/dfullingto… #MedTwitter #AIinMedicine #PhysicianAI #CME #ClinicalAI

@EvidenceOpen Thank you for sharing.

Research evidence takes an average of 17 years to change clinical practice. One in five validated interventions ever makes it to routine care. Doug Fullington, MD has practiced internal medicine for 25 years. He writes that OpenEvidence collapses those 17 years to about 17 seconds, in the exam room, with the patient watching the evidence arrive alongside him. The piece is not a product review. It is an honest account of what happens to a primary care practice when the distance between a clinical question and an evidence-based answer goes to zero. Including what the tool cannot do. Read full article: open.substack.com/pub/dfullingto…