THINGS MY PATIENTS TAUGHT ME — No. 1 She was ninety-one, and she'd driven herself to the appointment. That was the first thing that got me—the valet ticket still in her hand, her keys in a little coin purse with a broken clasp she kept snapping shut. Her heart was failing. Ejection fraction of twenty percent. I'd seen the echo before I walked in, so I came in the way I always do—fast, already talking. Medications. A possible device. The numbers, the odds, the plan. She let me get about thirty seconds in. Then she put her hand flat on the exam table and said, "Doctor. Sit down." I sat. "You people always come in like the building is on fire," she said. "I've had four cardiologists. Every one of you talks to the chart." She looked me up and down. "You're the youngest. And the fastest. I don't know if that's good or bad." I told her I'd take it as a compliment. "Take it however you need to," she said. "At my age I've stopped explaining myself." Then she got serious. "I'm not here to be fixed today. I know where this is going. I came so somebody would look at me while I said it out loud." So I put the folder down and I looked at her. She told me she was born in Poland. That she'd buried a husband and a son. That she still made her own soup, and that the soup was better than anything I'd ever eaten, and she was sorry but that was simply a fact. That she didn't want to die in a hospital with a tube down her throat. It took maybe eleven minutes. I didn't write a single note. I've spent thirty years learning to repair the human heart—the stents, the rhythm, the medications, the mechanics of it. Nobody taught me the harder skill, which is how to sit still next to a heart you cannot fix and not reach for a pen. She died about three weeks later, at home, the way she wanted. I wasn't her doctor that afternoon in my office. I was the person who finally stopped talking. And I've come to think that was the more important job. We assume the holy work is the fixing. A lot of the time, the holy work is just staying in the chair. Somebody in your life right now isn't asking you to solve them. They're asking you to look up. Blessings, P.S. She was right about the soup. I never got the recipe. I think about that more than I should. (Things My Patients Taught Me — No. 1. There will be more. They taught me more than I ever taught them.)

OIL SETTLED NEAR $85 AFTER BESSENT THREATENED ECONOMIC SANCTIONS ON ANY COUNTRY DOING BUSINESS WITH IRAN AS PART OF AN 'ECONOMIC D-DAY' DRIVE TO ISOLATE THE COUNTRY AND END THE WAR, WHILE ELON MUSK SAID SPACEX'S FIRST AI SATELLITES, POWERED BY NVIDIA'S TECHNOLOGY, WILL LAUNCH IN Q4 NEXT YEAR AND REACH SIGNIFICANT SCALE BY 2028.

Your Tesla gets better while you sleep Highlights from our upcoming Spring Release below — Visual updates (Model 3/Y with AMD chip) – Immersive full-screen vehicle controls when parked – Large playback controls & quick access to Recents, Favorites & Up Next in media player – Expandable Autopilot driving visualizations, with a smaller map in the top right for trip guidance Auto Shift Beta (Model S/X 2021 & later) Auto Shift (Beta) can now also shift between Drive (D) & Reverse (R) based on your surroundings, in addition to shifting out of Park (P). Vehicle settings > Pedals & Steering > Auto Shift (Beta) Audible @audible_com is now available as a native media app. Spotify queue & playback speed You can now sync your @Spotify queue across vehicles & devices & adjust playback speed. Hands-Free Trunk (Model S/X 2021 & later, upgraded Model 3) Stand still behind your trunk with Phone Key & the trunk will open on its own. Vehicle settings > Locks > Hands-Free Trunk Preview of Sentry Mode Recordings (select countries) Preview a brief clip of Sentry Mode recording on your phone if alarm is triggered by pressing & holding notification. Increased Regenerative Braking on Highway (Model S/X 2021 & later) At high speeds, your vehicle now allows for more regenerative braking, meaning more energy can return to the battery. Are we there yet?? Rear passengers can now see current trip details, time & temperature at the top of the rear touchscreen 👀

AI-driven R&D We expect AI systems to contribute more and more to AI R&D: that is, to be able to improve themselves. We’re researching techniques to ensure human visibility into and control over these systems.

A number of people asked what that Halloween night drive felt like. This became the last drive I’d ever take with my original heart. Thank you @Tesla and thank you @elonmusk for building the Cybertruck that saved my life. I wrote a song about the ride that night and have attached a small video clip of it here.

Do ChatGPT (OpenAI) and Claude (Anthropic) have any moats? There’s currently a lot of talk in the media and on X that frontier labs have no real moats, and that as they move toward public markets, they’re increasingly trying to secure them through regulation. Independent of what’s true there, let’s assess the moats they actually have today and the ones they could develop, using Hamilton Helmer’s 7 Powers. 🏗️ 1. Scale Economies: 5/5 Very strong for both, with a slight edge to OpenAI. Their sheer compute scale should create better unit economics through utilization, purchasing power, volume discounts and infrastructure optimization. But compute scale also matters on the product side. More compute means more resources to train models, experiment, run evals and improve products. Scale helps both margins AND model development. 🌐 2. Network Economies: Developing, potentially huge Does every additional user make the product more valuable to other users? Today: somewhat. Long term: potentially massively. Think agents communicating with other agents, businesses connecting their systems into the same ecosystem, users interacting through AI, and developers building apps for a massive installed base. More users → more developers/integrations → better ecosystem → more users. Anthropic has particularly strong potential here in B2B. OpenAI has that PLUS potentially enormous consumer network effects. This might ultimately become their strongest moat. ⚔️ 3. Counter-Positioning: Basically gone OpenAI initially attacked incumbents from a completely different position. Now OpenAI and Anthropic ARE the incumbents. Counter-positioning is increasingly being used against them: Open weights. Cheaper inference. Customizable models. Different business models. I wouldn’t consider this a meaningful moat for either anymore. 🔒 4. Switching Costs: Developing rapidly This could become enormous. Memory is the obvious early example. But take it much further: Your AI knows your entire history. Your contacts. Your company. Your files. Your preferences. Your workflows. Your agents. Your personal context. Everything is customized around you. At that point, switching AI systems could become incredibly painful. Today: moderate. Potential: massive. 🏷️ 5. Branding: OpenAI very strong, Anthropic moderate For most consumers, ChatGPT basically IS AI. That’s an enormous brand asset for OpenAI. I’m less convinced branding matters nearly as much in B2B. If an enterprise is spending millions and AI materially affects its productivity, it will eventually care much more about performance, reliability and economics than which logo is on the product. So: Consumer branding: very strong OpenAI moat. Enterprise branding: much weaker moat. 🧠 6. Cornered Resource: Weak today Compute overlaps heavily with Scale Economies. Could preferential access to NVIDIA GPUs or infrastructure become a cornered resource? Maybe, but probably not sustainably enough. Talent is another candidate. But if you’re paying extraordinary compensation to retain that talent, is it really cornered? A lot of talent will likely follow money and opportunity. One interesting potential cornered resource would be privileged regulatory or government access. If regulation eventually creates barriers that only gigantic incumbent labs can satisfy, that could become a very real moat against smaller competitors. That is one reason the regulatory capture debate matters so much. ⚙️ 7. Process Power: Unproven, potentially enormous Today, I don’t think this is proven. But self-improving AI could completely change that. Imagine: Better model → better AI researchers → faster research → better internal tools → better model → repeat. If frontier labs develop deeply embedded, proprietary AI-driven R&D processes that outsiders cannot easily replicate, Process Power could become enormous.

ALPHA DRIVE ONE Global Official Fanclub Early Bird 모집 D-DAY 🎨drawing by #JUNSEO# #준서# 🤍 #GEONWOO# #건우# 💙 #XINLONG# #씬롱# 🖤 #SANGHYEON# #상현# 🩵 📅 ~ 2025.10.26 11:59PM (KST) 🏎️ #ALPHADRIVEONE# #ALD1# #알파드라이브원#

ALPHA DRIVE ONE Global Official Fanclub Early Bird 모집 D-DAY 🎨drawing by #ARNO# #아르노# 💜 #LEO# #리오# 💙 #SANGWON# #상원# 💛 #ANXIN# #안신# 🩷 📅 ~ 2025.10.26 11:59PM (KST) 🏎️ #ALPHADRIVEONE# #ALD1# #알파드라이브원#