Sam Finlayson
@IAmSamFin
Followers
5K
Following
17K
Media
846
Statuses
10K
MD, PhD. Pediatrics-Genetics Resident @UWPedsRes Past: @harvardmed, @MIT_CSAIL, @Stanford Interests: Machine learning, Pediatrics, Rare Disease
Seattle, WA
Joined October 2014
"I’ve got you, Ron — that’s totally normal, especially with everything you’ve got going on lately." Who actually wants their model to write like this? Surprised OpenAI highlighted this in the GPT-5.1 announcement. Very annoying IMO.
67
15
555
No, look at *this* distribution of z-values from medical research! (329,601 z-values from Cochrane database)
11
48
577
Excited to launch an openRxiv partnership with the scientist-run AI review service qed science, the brainchild of @OdedRechavi
https://t.co/QOhJIsoDj9 1/n
openrxiv.org
Peer review is widely viewed as a critical aspect of biomedical communication. Ideally, it provides authors with feedback so they can improve manuscripts and gives readers, particularly nonspeciali...
2
21
59
Injection of patient-aligned AI bill reviewing will create new incentives that will generate new market forces. Opportunity: Regulations that widen the aperture for AI-driven patient-aligned value/$ perspective. CC @IAmSamFin @zittrain
A guy just used @AnthropicAI Claude to turn a $195,000 hospital bill into $33,000. Not with a lawyer. Not with a hospital admin insider. With a $20/month Claude Plus subscription. He uploaded the itemized bill. Claude spotted duplicate procedure codes, illegal “double
2
5
13
You've all seen this viral graph on social media Shows healthcare administrators have increased 5000% since the 1970s! Hey, even @TheLancet published it In the past, I've tried to replicate it and failed So this time, I asked @bhrenton to do it right
5
37
122
📣 Announcing the 7th Annual Conference on Health, Inference, and Learning (CHIL) happening June 28-30, 2026 in Seattle, WA! 👉 Call for Papers is up at https://t.co/8Pp68WYlLV... ⏳ Submit by February 4, 2026
1
5
9
What's striking for me about this take is that it is almost precisely the opposite of what I would consider the moderate view. (Namely, that even if AI progress stalled today, there are tons of low hanging fruit that could yet be plucked using todays models).
I'm desperate for a worldview where we agree both are true: - current AI is slop and the marketing is BS, but - staggering AI transformation (including extinction) is 5-20 years out, this may not be good by default, and thus merits major policy action now
0
1
7
New Roche sequencer being tested for same-day genome interpretation in neonatal ICUs: "In the case of blood samples that arrived at the laboratory by 7 a.m., we obtained an interpreted report between 2 p.m. and 4:30 p.m. the same day." https://t.co/1wt1Z10QjH
1
48
132
This is why I’m generally anti-supplement — whatever benefit you think you’re getting from the “active product” is probably dwarfed by whatever mysterious stuff gets rolled in along side it
@ConsumerReports for context: taking just one serving of one of these powders would expose you to 45% more lead than the average american gets from everything they eat/drink in a day COMBINED that means daily users of these powders could easily exceed the FDA’s suggested daily dietary lead limit
1
1
3
Benchmarking autosomal recessive disease prevalence estimation from allele frequencies against newborn screening data | medRxiv
medrxiv.org
Accurate estimates for the prevalence of rare congenital diseases are critical for understanding disease epidemiology and enabling drug development. Prevalence estimates can inform public health...
0
3
5
I just learned Claude's new code interpreter mode has a /mnt/skills/public/ folder full of prompt instructions and Python utilities for creating and manipulating pdf, docx, pptx, and xlsx files - and you can ask Claude for a copy and learn a TON about working with those formats
35
124
2K
Gonna get slammed as the Luddite doctor but, “the personalized health AI future we all want” is… Daily (?) thyroid labs but without lab economics of scale?
The personalized health AI future we all want, and so many are rushing to build, depends on solving the blood testing bottleneck. SiPhox is a decade ahead of anyone in delivering a Nespresso experience for home testing. 🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸🩸
0
0
3
A brief history of what we know about de novo germline mutation in sperm and it's relevance to the paternal age effect in autism. 🧵1/N
3
23
71
Hard not to feel touched and inspired when someone with a noble vision spends a full career focused on achieving exactly one thing, and then does so. Wilson’s Disease to enter the newborn screen after a three decade effort from a wonderful clinician and man
2
1
5
This is an excellent history of LLMs, doesn't miss seminal papers I know. Reminds you we're standing on the shoulders of giants, and giants are still being born today.
12
114
693
I’ve been obsessed with the @NEJM CPCs since I was in grad school. Now, with @tabuckley_, it’s surreal to see @NEJM publish the first AI differential diagnosis in the 100+ year history of the series, generated by our AI system Dr. CaBot, alongside the human expert’s.
7
74
286
If you've been trying to figure out DSPy - the automatic prompt optimization system - this talk by @dbreunig is the clearest explanation I've seen yet, with a very useful real-world case study https://t.co/w6uXcGshwI My notes here:
simonwillison.net
I've had trouble getting my head around DSPy in the past. This half hour talk by Drew Breunig at the recent Databricks Data + AI Summit is the clearest explanation …
19
154
1K
Today is the 54th anniversary of the first CT scan 🎉
13
300
2K