Internal Medicine physician interested in healthcare, technology, and creative pursuits. Leading product @PatientsApp Ex-data @oscarhealth and @dropbox
New York, NY
- Tweets 9,603
- Following 1,638
- Followers 2,000
- Likes 12,125
this is very cool
Introducing Synthetic Hospital: an open, fully synthetic longitudinal EHR benchmark with verifiable ground truth!
1,268 patients, 5,602 encounters, zero PHI. Physicians could not reliably distinguish its charts from real ones.
📄 arxiv.org/abs/2609.30027
💻 github.com/sparkcpark/synthe…
✍️ sparkcpark.github.io/posts/f…
1
1
20
9,103
i think everyone should have a side project they are hacking on for fun that solves a problem that's only for them
not to scale, not to grow, just truly to tinker. play with, and build things for yourself
it's not that serious
some projects i'm working on:
1. my own ai fitness coach (exercise, nutrition, sleep - sorry longevity bros it's all i need)
2. a letterboxd replacement for movies and tv shows that's nice to use
3. a hospital scheduling tool (building for a friend) that can also find shifts to swap (building for me)
4. a personal financial advisor - that helps me keep track of money, loans, investing, etc. (i didn't want to pay for a financial advisor after residency which some people recommend)
7
1
28
1,222
this looks very interesting and i’m always excited when companies are focused on prevention in a focused way (ie not 100+ biomarkers like function)
the only strange part about this is that CAC test is not really that useful in most people. it’s really for specific populations where you’re borderline deciding to start treatment. so offering it as a general screening scan imo doesn’t do much
not to mention it will only show hard plaque (not soft plaque like a CCTA) and in many cases you’re better off using risk calculators and labs to make decisions about treatment
selling a scan like this with nice data and a good experience feels very similar to full body MRI companies selling the experience vs something that is actually clinical useful
Today, we’re officially launching @HalePrevention with a film called “Grow Old With Them”.
You won’t see us founders in it. We wanted you to meet one of our customers, and hear his story in his own words.
It captures why we built Hale better than any founder introduction could.
3
4
1,804
cc @DrSiyabMD curious about your thoughts on this since this is your area
55
with all the personal agents popping up it makes you think if specialized vertical agents will have a place
the argument that dedicated harnesses for specific tasks is needed seems to not be holding up as we see more capable general agents that can do it all
i still think that for many avg consumers there will be some pull for having specialized apps/agents for bigger areas like health, fitness, and finances but less so for things that a general agent can likely do well (eg fashion, home tasks)
1
5
484
more full body scans and imo likely unnecessary biomarkers
totally ok to want data about your body if you can afford it but the outcome is usually just as packy notes here - need to exercise more, eat healthier, and sleep better
my only concern with fully body scan companies is that they will take up radiology resources from the rest of the market (similar to what we say headway and other mental health aggregators do) and they will drain time and resources from the system for following up and doing biopsies on incidental findings
this is a good health tech business but likely not a good patient outcome / healthcare system outcome business
the stories that make the headlines will be ones where things are caught early but the medical waste likely won’t be reported on and i wish more people talked about that
Neko Health is as good as advertised.
I got to be the 4th person to preview the new NYC location yesterday. The whole experience, from the moment you enter and head down the stairs, was dialed in. It felt like the future.
Beautiful combo of design, technology, and people. Got a bunch of scans, bloodwork, and blood pressure done in ~30 minutes and as soon as it was done, they took me into a room to explain the results, which were all ready that quickly because there’s a lab on site (I won’t spoil how the blood gets there, maybe the coolest part).
Luckily, I’m basically healthy but need to get my fat ass in shape.
It’s wild that it costs $499, and they’re going to keep squeezing more tests into that price. They were nice enough to give me this one on the house but I already paid for my next appointment in 12 months.
10/10 would recommend.
3
1
5
1,257
i will say the auto crons instinct sets up do make it feel like magic
still running both that and @bot
186
this isn’t the best rebuttal for AI not being good
agree that ofc it can’t do a physical exam (or renew the tele for that matter 😉) but it’s not a fair assessment to test it on things that aren’t there
this is likely solved by a harness issue that looks for more pertinent negatives - you can’t look for all of them but you should look for the right one
less of a model issue, more of a harness issue
The pattern is this. The assessment ruled heart failure unlikely because the NT-proBNP was low. The patient was obese, which lowers that number, and the JVP was elevated. The JVP was nowhere in the chart, so the model read missing documentation as a normal exam.
Not a hallucination. Every fact it used was in the chart. The error was the inference from silence, which is why a stronger model reading the same chart makes the same call.
2
716
sick burn by Instinct
Introducing Muse, your personal AI agent from Meta that gets things done across every part of life.
Download the Muse app and get started: Muse.ai
2
542
function health?!? nooooooo
1/ we’ve built a lot of connectors to make it easier to integrate Muse into your life. Muse runs in a secure VM and we don’t use your data for ads or anything other than to make Muse great for you. you can connect Muse to Gmail, Google calendar, Outlook, Plaid, Opentable, Google Docs, Spotify, Function Health, Withings, Tailscale, Peloton, and more.
2
1,280
great analysis here that shows unfortunately that when systems can be gamed, they will be
i worked on surprise billing for a bit at Oscar and it was always such a tough spot for everyone - patients getting hit with bills, providers not getting reimbursed enough, payers trying to figure out what the right reimbursement amount is. IDR was supposed to work to fix that but the data looks like it’s become a game (it was always supposed to be a game but one side is clearly winning more than they likely should)
it’s so tough to create a program that looks out for patients and create fair reimbursements but hopefully some of this data can be used to make improvements in the future
of course - centralized rates would solve a lot of this but that also has its pros and cons
chir.georgetown.edu/spending…
4
644
so many applications of these meme
Hormone tracking for women with a smart earring! Sounds fantastic, but the story of Phira also shows what's problematic these days with new companies developing wearables.
I scanned through both their pages on the website; one is about how the device works and the other is the science behind it. I couldn't find the technological explanation I was looking for.
The first page is about what they expect to add to women's health and the science page is more about the science behind the reason for tracking hormone levels. The only paper is a presentation they gave at an IEEE event. The clinical pilot study they mentioned has no citation, so we cannot look up its data.
I'm sure they know what they are doing and maybe this whole secrecy is around protecting their IP and patents, but the medical community operates on the concept of evidence-based medicine.
If you can prove that your technology works in studies, I'll be your biggest supporter. Until then, I always have Theranos PTSD no matter the company or the technology.
1
4
1,117
pro tip for using Instinct on mac via @raycast
since there is no dedicated Instinct app, you can set up a quick link to open your Instinct iMessage thread to quick jump to it from anywhere
261