Sebastian Caliri

Everyone knows about serious problems in American healthcare: $1.8T in federal spending, burned-out doctors, endless phone trees, incomprehensible bills. No-one has a viable plan to fix this mess. And rather than helping, our policies have effectively banned AI from medicine. We have made our very best hope for relief illegal. I spent the past few months talking with policy makers in DC, state secretaries of health, and the technology community about this problem. I spoke with doctors about their experiences and ordinary Americans about their challenges with the healthcare system. I also spoke with investors about the barriers to investing in healthcare AI that makes life better for patients and doctors instead of pouring more dollars into up-coding tools. From these conversations I synthesized a set of policy reforms at the federal and state levels that will not only legalize healthcare AI, but also attract resources to the space and get the very best innovators focused on what is an existential problem for the United States.

Few venture dollars are deployed to clinical AI *at all* right now, because most of the valuable use cases have no regulatory pathway at FDA and at the state level. And to the extent CMS sets a lot of the incentives in American healthcare - I mention Congress but really a stand in for federal government. If you create the right incentives industry can get to work. The most exciting incentive structure I ever saw relevant to rural health was the geographic DCE model, which was canceled at the start of the Biden admin.

@SebastianCaliri There is no market structure. Venture dollars are not deployed to highest quality lowest price solutions. They sell to health systems and payers, the highest cost players.

Congress is in bipartisan agreement to welcome more foreign-trained physicians to our country. Congress is silent on or opposed to American physicians using the best American technology to extend low cost urgent, primary, and specialty care to rural Americans.

The full deck on Sid’s cancer approach is here: sytse.com/cancer/ Worth a read. Raw data for download is also available and linked in the deck

Evidence-based medicine is a blessing of the 20th century. Evidence-based medicine is also a curse of the 20th century. Medical interventions are studied through randomized controlled trials and those interventions are assessed for efficacy and safety on a population level. But no individual quite matches some blended average of every trial participant. Rather, everyone's biology is unique. Sid Sijbrandij just presented the story of his cancer journey at OpenAI forum. When Sid ran out of evidence-based treatment options, he didn’t accept the boundary but rather began treating his cancer like an engineer: - multi-omic tumor profiling at extreme depth - N=1 drug development (vaccines, TCR-T, radioligand therapy) - parallel treatment strategies - continuous measurement (ctDNA, single-cell, immune state) and refinement Rather than protocol-based care, Sid built a learning loop. Maybe the future of medicine in a world where gathering and interpreting data gets cheaper and cheaper looks more like a loop. Thanks for sharing @sytses and @jacobjstern !

@kane @SFBART The only thing scarier than 24th St Mission BART is 16th St Mission BART.

@SFBART Wow that’s great, what are the Downtown Oakland and Mission BART stations like?

This was Downtown Oakland to the mission district in SF.

You probably have seen the headlines on the latest paper bashing AI. Our response to the Nature Medicine paper claiming ChatGPT Health under-triages 51% of emergencies is now on arXiv. Short version: the evaluation format drove the failure, not the models. Link below 👇

@arijitchakrav I judge the work on its merits, not on who wrote it. I found the arguments about how the paper does not align with real world conditions to be compelling.

@SebastianCaliri The paper you’re referencing is a peer-reviewed nature medicine paper. The rebuttal (99%) is from a blog post by an OpenAI researcher. Do you weight those two pieces of information equally in your mind?

Remember the paper saying ChatGPT Health undertriaged 52% of medical queries? Turns out the study methodology was massively flawed. In real world conversations, GPT-5.4 correctly recommends immediate care in emergency cases more than 99% of the time. Healthcare AI isn't perfect and there are open questions about how to evaluate LLM-based tools. But always worth remembering that there are groups that have an incentive to produce a sensational, AI-skeptical headline by cloaking suspect science in academic authority.

@jluan Not good. They found out about the 90-minute synthesis. CC @hyperindexed

apparently this is what the chinese AI ecosystem thinks our grand american AI master plan is! (this is not a joke! forwarded to me by an attendee at a real chinese ai conference!)

Thoughtful and more carefully-framed article from Karan Singhal here: x.com/thekaransingha…

We wrote the first check into Quince and never shared our thesis publicly At the time, D2C was radioactive and 'affordable luxury' sounded like an oxymoron. Today they raised at $10B. We couldn't be more excited. Here's why we believed — and why we think it's still early! 8vc.com/resources/quin…

What is this man in the boat doing? And why are our tax dollars paying for it? Good questions, but it was a beautiful morning over San Francisco Bay.

Pricing for deflation (=reducing cost per health outcome ) is largely a policy choice. I think ACCESS strikes a good balance where in theory health systems can also win in the process. That is contingent on health systems being constrained on physician hiring (true of at least 8-9 health systems CMOs I polled) and wanting to shift lower complexity cases / encounters to supervised AI to free up time. I hear anecdotally that new attendings simply elect to work fewer hours than retiring attendings which will exacerbate demand on physician time

Cardinal fallacy is assuming the AI sponsored care will be free/low cost. Reimbursement models covering AI are already being considered. (Read up on @SebastianCaliri) Health systems will lobby and leverage this to insure money keeps flowing their way. (It actually fits nicely into their thesis - deliver healthcare without having to pay physicians.) Their goal isn’t to make healthcare costs more affordable- it’s to secure market share, contracting power and make sure money keeps flowing their way. Your bulwark against this are physicians - if we can ever get our collective act together.

Americans now pay more for healthcare than housing. The **only** solution to this is to legalize AI providing medical care in all 50 states Doctors will certainly still be in the loop (similar to FSD) but this is the only solution that collapses the cost of care Many lives will be saved.

Last year there were about 70 adults in OCIA (new converts). That was already an all time high. This year there are 100 in OCIA. This is on a base of ~4,000 registered parishioners. Catholicism in SF is growing stronger!

At least 1/3 of California's tax base comes from high earners in the technology sector. "Tech bros" are footing the bill for every social program and benefit in this state. Gen Z / Millennial tech workers in California (who are mostly not billionaires capable of sophisticated tax games) get soaked on income, while paying property taxes that are 2x-3x as high as our Gen X and Boomer neighbors due to Prop 13, while paying for private school because of our unaccountable public schools, all while getting insulted by politicians like Steyer. If Steyer was serious about helping working families, he would be trying to build a coalition including all types of Californians. Instead he just puts down some of the people the state needs the most to flourish. Message Received

I would happily invest in AI companies providing accurate medical advice as certified by a government authority - if the government would create such an approval process! The proposed ban is lazy. The harder thing that is constructive for society is defining what accuracy and quality looks like and creating a way for companies to show they meet that bar.

@practiklyperfct 11:30 am every Sunday!

@SebastianCaliri I will next time I’m in town.

Be a contrarian and come to mass at St. Dominic’s.

I’m extremely honored to be nominated by President Trump to serve as Director of the National Science Foundation.   Heroic scientists have always challenged consensus to advance the frontiers of knowledge. Recently, many institutions have weakened academic freedom and lost the trust they once enjoyed. Yet across our country, a new golden age of discovery is dawning. Information is open source and debate is public.   The marketplace of ideas is not an efficient market. Finding and funding independent thinkers and builders has taught me to eliminate bottlenecks and favor rigorous science that replicates.  Private funders are developing frontier models and useful technology. Government should take bigger financial risks to pose and answer deeper questions.   NSF’s scientists and staff have built something worth strengthening. Working together, scientists, engineers, investors, research institutions, and businesses can support American genius, enhance national security, enrich our economy, and improve our quality of life.   Entropy is on the march and China is not waiting.

@maybeltr This year, at least 17 states will make it impossible to use SNAP for candy and soda. HHS has been encouraging states to implement this waiver program and hopefully more will get on board. Anecdotally, various food lobby groups hate this policy, but it’s finally getting done.