Microsoft's AI can diagnose NEJM cases at 85% accuracy, but real hospital charts are too messy for today's LLMs. The bottleneck isn't reasoning—it's context engineering that can parse fragmented, stale EHR data into something models can actually use.
When You Upload Your Medical Records to AI, Who’s Actually Protecting Them?
Forty million people ask ChatGPT health questions every day. This week, OpenAI and Anthropic made it official: connect your medical records, sync your Apple Health data, let the AI see your full health picture. The product pitches emphasize encryption, privacy controls, and promises not to train on your conversations. Here's what they don't mention: your … Continue reading When You Upload Your Medical Records to AI, Who’s Actually Protecting Them?
A Year of Talking to Computers
I shipped four iOS apps in a month without being an engineer. AI collapsed the distance between idea and reality. The bottleneck moved from building to judgment.
Your Gut is a Value Function
Trust your gut" isn't mysticism—it's a value function trained on years of real outcomes. AI can't replicate it because current training uses short-term feedback on curated datasets, not the messy, long-horizon reality that shapes human judgment.
Strategy in the Age of Infinite Slop
AI won't replace strategy consultants because corporate decisions yield sparse, delayed feedback with no A/B test. While LLMs generate options brilliantly, only humans can judge which will survive contact with reality years later.
The New Computer in the Clinic
Healthcare is moving from "point-and-click" EHRs to "intent and oversight." LLMs can finally read clinical narratives, but they need deterministic guardrails to convert physician storytelling into safe, structured actions patients can trust.
AI and the Prepared Mind: Engineering Luck in Drug Discovery
AI's "Logic Engine" designs perfect molecules in silico, but most still fail in Phase 2. The gap isn't chemistry—it's understanding human biology. Dr. Eng's GLP-1 discovery came from embodied clinical context AI can't replicate yet.
How AI Gets Paid Is How It Scales
AI agents will scale in healthcare when they create a labor dividend—either eliminating admin overhead or letting each scarce clinician produce more. Reimbursement models must shift from human minutes to AI-driven outcomes.
When AI Meets Aggregation Theory in Healthcare
Epic isn't a true platform—it doesn't pass Bill Gates's test where ecosystem value exceeds the company's own. With IAS rails and consumer AI assistants, healthcare's first real aggregator could finally own demand, not just supply.
America’s Patchwork of Laws Could Be AI’s Biggest Barrier in Care
State AI rules are regulating software like a risky human, creating 50 versions of compliance. A federal framework could separate assistive from autonomous agents, letting FDA handle high-risk use while safe harbors accelerate adoption.
The Gameboard for AI in Healthcare
Healthcare AI is stuck in "middle to middle"—helping clinicians draft and summarize, but not prescribing. Level-5 autonomy requires moving from probabilistic language to deterministic actions with verifiable plans and safety gates.