Blanca

The idea for Demis's work on memory and imagination (a Science top 10 breakthrough of the year) he had on his honeymoon in Italy. This plus David Silver’s thinking of AlphaZero on his honeymoon in Sri Lanka makes me think we should also go on more honeymoons. 8/

@JTLonsdale @mcuban Building out hospital-level infra is a non-starter for surgery centers and rural sites, but care gaps won't be filled by hospitals. 'Question, delete, simplify, accelerate, automate' is necessary for healthcare abundance that actually serves patients.

@mcuban Yeah

Why aren’t any of these at risk hospitals publishing their full accounting so everyone can see where they spend their money ? All but one group of hospitals that I have looked at potentially investing in, spend so much on consultants and fees that it’s no wonder they are at risk Plus, I have NEVER seen an industry that is worse than hospitals when it comes to buying medications and items like implants, screws, other devices. They overpay for everything. And then when you show them how to save money, their “supply chain” employees resist any change. They are so set in their ways, it’s a shock more don’t go out of business. Prove me wrong.

@naval llms and agents knowledge surface is still spiky, so uncovering failure modes is fun too. To your point, much easier to suss out gaps when you know what you want to build.

Vibe coding is more addictive than any video game ever made (if you know what you want to build).

@mcuban Building out hospital-level infra in ASCs is a non-starter for exactly the reasons you cite, but care gaps won't be filled by hospitals. 'Question, delete, simplify, accelerate, automate' is necessary for a healthcare abundance paradigm that actually serves patients.

Space — the final frontier. Our new National Space Agency begins operations today. The global space industry is growing fast. While Singapore may not have launch sites, we have strengths in specialised, high-value areas — from satellites to advanced manufacturing.

@beffjezos AI for healthcare can also totally enrich docs. Capturing more cases, providing more care/increasing demand, improving outcomes. Upside can be shared in a healthcare abundance paradigm.

Narrative violation. AI for healthcare will make doctors more efficient, lower costs, and increase demand for health services

@rtnarch Also on the table is to not replace anyone, but decouple care delivery from physician time and geography, give existing docs 10x reach

Replacing BOTH hospital administrators and radiologists with AI is probably coming.

@martinvars Physicians don’t scale, but procedures do—and procedures can be automated. Fed gov tailwinds are pointing in this direction. Physician scarcity is not going away. Physician geographic density won’t change soon. Decouple care delivery from geography and doc scarcity.

@martinvars “Human in the loop” = doctor on the hook. Access crisis is not going away via training or paying $$$ for specialists to move rural. Have to share tech upside with physicians and 10x reach with automation and autonomy.

The US healthcare system is currently spending hundreds of billions annually on administrative bloat, yet millions of Americans still live in care deserts. We have a severe physician shortage that no amount of medical school admissions can fix in the next decade. Yet, almost all the capital in healthcare AI is flowing into 'copilots'—tools that just add another layer of software for an already exhausted, burned-out doctor to manage. At Certuma, we realized this is the wrong approach. You don't solve a catastrophic labor shortage by giving the missing labor a better spellchecker. You solve it with fully autonomous clinical AI. We are building autonomous doctors designed specifically for environments where the alternative is no care at all. To do this, we are putting our AI through the exact same rigorous regulatory scrutiny as a new drug or medical intervention. No shortcuts. No 'human in the loop' safety nets to hide behind. If we want to actually increase the supply of healthcare and drive down costs, we have to stop treating AI as an administrative assistant and start treating it as a highly regulated, highly scalable medical intervention.

@zakkohane A social science framing of the nasty moat is illegible v legible systems (Seeing Like a State). Healthcare at the population level remains largely illegible.

Let's hope that healthcare does not fit perfectly into the nasty moat described in this intriguing Uber-for-VC ;-) agentic Y-combinator-fast-follower/cloner effort.

Applied Intuition CEO @qasar says the market for physical AI is "way, way bigger" than the market for white-collar AI: "I used to be at Y Combinator. I was the COO, ran the firm, and funded lots of interesting companies. And one of the analogies I used to use to help founders understand market potential and size is: I grew up in Detroit. You're sitting in the Detroit metro airport at a gate, and you look around. How many of those people are using Claude Code? Frankly speaking, not many." "But how many of those people drive? How many people work at construction sites? How many of those people ride in buses? How many of those people serve in our armed forces? The point is: a much, much larger group." "The market for physical AI is way, way bigger. Purely because the surface area is much bigger."

@ColtonOrtolf I agree paying specialists more has not worked to bring them to rural counties. There are tailwinds from gov, tech, and med to fill Bucket 3. Physical AI is mature enough to make a dent there.

Supply chain attacks like the currently breaking axios, litellm and xz are only going to be more commonplace in the vibecoding world. The entire premise of vibecoding is “I don’t need to understand the code” happens to also be the entire premise of a supply chain attack.

American physicians are remarkably pro-AI, and getting more supportive each year. 76% of doctors believe AI can help their ability to care for patients. And 70% believe that patients’ use of general-purpose AI chatbots for health information is positive / or has no impact. AI will not become a part of American healthcare without buy in from physicians (i.e. individual physicians, not the AMA or other societies, that don't necessarily speak for docs). Misguided ideas like the NY state bill would limit patient access to these tools. But most American doctors understand that banning data centers is bad for patients.

@SebastianCaliri @martinvars Thanks for your time talking through this w me on the walk from H&V. That 76% are the champions who will actually bring innovation to bedside. Technologists need to provide protections, be physician-friendly, and share upside for it all to work.

Excited to release two new AI papers. First, we report what we believe to be the first truly multi-modal visual language model for cardiology. Second, we report surprising findings related to how visual language models reason. arxiv.org/abs/2603.22179 arxiv.org/abs/2603.21687

@fchollet LLMs are very spiky right now

One of the biggest misconceptions people have about intelligence is seeing it as some kind of unbounded scalar stat, like height. "Future AI will have 10,000 IQ", that sort of thing. Intelligence is a conversion ratio, with an optimality bound. Increasing intelligence is not so much like "making the tower taller", it's more like "making the ball rounder". At some point it's already pretty damn spherical and any improvement is marginal. Now of course smart humans aren't quite at the optimal bound yet on an individual level, and machines will have many advantages besides intelligence -- mostly the removal of biological bottlenecks: greater processing speed, unlimited working memory, unlimited memory with perfect recall... but these are mostly things humans can also access through externalized cognitive tools.

States passed 33 healthcare AI laws in 2025 while the feds went hands-off. Healthcare AI startups now face 50 different definitions of 'medical advice,' 'clinical decision support,' and 'patient data.' Same AI tool, 50 different compliance frameworks. We've fractured healthcare innovation into state-by-state silos. Big players with legal armies win by default.

Someone in the admin should look into this

This is my #1 piece of comms advice for startup founders.