the single most insightful article i've ever read on the history and state of modern medicine
@MidwesternDoc
open.substack.com/pub/amidwe…
246
i think the foundational point here is exactly right. and i think it’s pretty important to get this right, right now. nice that it aligns with my gut all year, but i think it’s becoming increasingly self evident
$PLTR $HIMS $DUOL $SPOT $RXRX $NAUT $TEM can all grow free cash flow per share without growing CapEx.
Wall Street thinks the AI winners are the companies spending the most on GPUs. I think it's the ones that need them least.
1/ The frontier labs and hyperscalers are in an arms race. Every model generation needs more chips, more power and more data centers. Hundreds of billions go in, and what comes out is generic intelligence: a model that knows a little about everything.
2/ Generic intelligence is commoditizing faster than any technology in history. The same capability costs a fraction of what it did a year ago, and open-weight models trail the frontier by months, not years. You can't build a moat on something that gets 10x cheaper every year.
3/ What doesn't commoditize is proprietary data. A generic model has never seen a defense agency's operations, millions of patients' outcomes, billions of learner answers or a wet lab's phenomic maps. That data can't be scraped, bought or synthesized. It's generated by running the business.
4/ I call the loop Ontology Velocity: proprietary data → open-weight model trained on it → better outcome for the customer → more usage → more proprietary data. The faster it spins, the wider the moat.
5/ The unit economics are the whole thesis. Fine-tune an open-weight model on data only you own and it beats a generic frontier model on your customer's problem at roughly a 90% discount. Cost per outcome falls 90% while value delivered rises, so value per dollar goes from 1x to more than 10x (Exhibit 1).
6/ These companies don't fund the arms race. They're its customers. Every time the labs spend billions pushing open-weight models forward, these companies get the upgrade almost for free and run it on data nobody else has. The scaling laws work for them, not against them.
7/ That's how FCF per share compounds without CapEx rising alongside it: better models make the product more valuable, which brings pricing power, retention and operating leverage.
8/ $PLTR: The Ontology maps each customer's operations into decision-ready data. AIP turns any model into a decision engine on top of it. Palantir doesn't need to own the model.
9/ $HIMS: Every patient adds preference, adherence and outcome data, which powers personalized care a generic chatbot can't deliver.
10/ $DUOL: Billions of answered exercises show exactly where each learner struggles and what fixes it. The model is a commodity, and the learner graph is the asset.
11/ $SPOT: Listening behavior at planetary scale. Recommendations, AI DJ and discovery run on a taste graph no competitor can buy.
12/ Biology is code, and here the thesis gets more extreme, because the data isn't on the internet at all. It's generated physically, one experiment at a time.
13/ $RXRX builds phenomic maps in its own automated labs. $NAUT measures the proteome at a scale that produces data that doesn't exist yet. $TEM links multimodal clinical and molecular records to outcomes. No amount of GPU CapEx recreates these datasets.
14/ The caveat: the bio platforms are earlier on the curve. The data engines are being built now, and FCF comes later. Same thesis at an earlier stage, with more risk and more asymmetry.
15/ The framework: don't own the companies spending the most to train models. Own the companies whose proprietary data makes every new open-weight model worth more to their customers.
28
great article by $bfly ceo
For over 15 years, @ButterflyNetInc has been working to make ultrasound more accessible. What we didn't fully appreciate at the beginning was just how much bigger the opportunity could become.
Ultrasound-on-Chip™ is no longer just about making imaging portable and more affordable. It's enabling new models of care delivery and becoming a platform for innovation, connecting AI to the human body in ways we're only beginning to imagine. I share more on this next chapter in my latest article: linkedin.com/pulse/era-ultra…
At #HLTHUSA 2026? Join me Monday, 11/16 at 10:30 am PT on the Revv Stage as I dive deeper into The Era of Ultrasound Abundance.
$BFLY #Ultrasound #BCI #Neurotech #tomography #midjourneyscanner
1
1
171
i’m new here but this tracks with my experience
AI is the only theme where:
- a lot of capex = bubble
- less capex = it's dead
- unchanged capex = it's slowing down
- increased capex = it's not as fast as it used to be
- stocks up = overextended
- stocks down = still overvalued
- stocks giga down = too many risks
- good earnings = it's peak of the cycle
- mediocre earnings = it's over
1
27
how is it that this can freely circulate the internet, yet mere questions about the origin of covid or the safety of a new vaccine could not? oh, it has to do with which party is power. got it
At an event with Kara Swisher, Scott Galloway called for the execution of Elon Musk for being a “traitor to the west” live on stage.
1
96
what the hell
If you sponsor this podcast, calling for the execution of Elon Musk, I will never use your product… and will buy from both of your top two competitors
This is unacceptable and dangerous.
This a dog whistle to murder someone because you disagree with them — and this is the same guy who joked about the assassination against President Trump.
Fuck this piece of shit.
26
this should be studied
Most employment is sustained by social ties. Looking reasonable helps preserve those ties, and pessimism is the easiest way to look reasonable
That worked in a world where outlier success was rare
Now outlier wins happen every day, which is why pessimistic people are losing so much
27
what is the term when you take
money from one investor to pay the gains you owe to another?
BREAKING: The US government borrowed a record $839 billion this week via its sales of Treasury bills and notes, both to repay holders of maturing debt and to raise the additional capital needed to finance the federal deficit.
31
excellent overview on $naut
200k in revenue, 400m market cap, and I still think $NAUT can get to billions. Went through the Q3 call and put it all in a video. What they actually do, how much cash they have, the dilution risk, and what I want to see next. Micro position for me today. Not advice, I'm long.
1
184













