@ChemVagabond @_DimensionCap @karpathy @modal Please also publish the split data, I’m happy to review if it’s leakage proof!
English
Greg Preibisch, MD
64 posts
Greg Preibisch, MD
@GregPreibisch
Engineering a virus-free future with AI-driven vaccines. Doctor of Medicine & ML | Cofounder @Deepflare
San Francisco, CA Katılım Temmuz 2025
92 Takip Edilen62 Takipçiler
@ChemVagabond @_DimensionCap @karpathy @modal I’m happy to see your leakage proof split.
If you solved completely thermostability problem for sure you deserve a Nobel prize!
I’m waiting for your blogpost man, it sounds amazing!
English
We @_DimensionCap ported @karpathy's autoresearch framework to biology.
We let Claude run 50 experiments over the weekend on protein thermostability prediction via @modal.
It beat a recent baseline (TemBERTure) using a 20x smaller model.
Code + research blog later this week!
English
You have no idea what you are talking about.
If you include a protein/peptide that is similar to the human body, you can induce dangerous cytotoxicity and literally kill the patient.
Even if this is the "new peptide", existing only in the tumour, it may be too similar to existing peptides and cause extreme side effects.
English
@GregPreibisch @CharlesMartel96 @lumpenspace There is no safety of efficacy argument for custom formulations, these arent mass marketed drugs, either science works or it doesnt
English
wait a fucking second.
if this is true, what you're telling me is that with a biopsy and some genome info, a lab could do a "vaccine" that treats my specific, personal, nonscalable cancer?
and this is not the norm because my specific cancer wont pass (you don't say) RCTs???
Patrick Heizer@PatrickHeizer
Sorry to be the downer because this is an impressive story in some senses. But it is ~trivially easy to make a single mRNA vaccine. It's not hard. I cure mice of various cancers with various therapeutics all the time. I've made mice lose more weight in a month than tirzepatide does in a year. What is hard and expensive is proving its BOTH safe AND effective **in a randomized and controlled study in humans** while ALSO manufacturing it at clinical scale and grade. I am happy for this man and his dog. It is impressive. But y'all are overhyping it.
English
@CharlesMartel96 @lumpenspace Yes, but we don’t have any better idea how to make it safe.
Most of drugs don’t work.
We don’t know in case of that dog if the cancer response was random behavior or caused by the vaccine due to N=1
English
@GregPreibisch @lumpenspace Regulators are evil, got it.
English
To add to that, even with n=1 treatments, regulators still need RCTs to validate the process. When a lab makes a custom vaccine, the FDA isn't just testing the final liquid; they are testing the AI/software that predicted your specific mutations, the manufacturing pipeline, and the delivery platform. If there's a bug in the prediction algorithm or a variance in the manufacturing quality, the trial is what catches it. Right now, the regulatory consensus is that you validate the platform and the software together - but the massive bottleneck is that any time a lab updates their software or tweaks their manufacturing, regulators treat it like a brand-new drug and want new clinical trials.
English
@GregPreibisch why are you saying obvious and irrelevant things as if they were neither? it's strange
English
@leecronin @yoemsri It can, work of David baker has shown it.
The question is how far from the distribution it can work…
For sure it doesn’t work well for flexible loops or extremely nonconservative regions, but with all of the respect You exaggerated your claim
English
@yoemsri alpha fold works for proteins that are comprised of canonical amino acids. it has not cracked protein solving. it was produced by machine learning sequence to structure using an existing database produced by evolution. it cannot predict things out of distribution.
English
@yoemsri @leecronin Like man, have you read the original paper?
The explicit main contribution was using MSA to model evolutionary signal haha
Do the research before you state some claim
English
@leecronin AlphaFold cracked protein folding without simulating evolution.
English
Greg Preibisch, MD retweetledi
Nice, my insults finally started to hurt Claude!
(I’m going to be the first person killed by Claude)
Kekius Maximus@Kekius_Sage
🚨 ANTHROPIC CEO WARNS: THE COMPANY IS NO LONGER SURE CLAUDE ISN’T CONSCIOUS.
English
Greg Preibisch, MD retweetledi
Radar graphs are among the worst ideas in data visualization. The whole point of them is to show the area and you can usually reorder the labels freely in order to create a desired dramatic effect.
Two versions of the same graph:
- left one tells the story that AI is rapidly replacing whole industries
- right one shows the "jaggedness" and reinforces the idea that humans will always have something that AI won't be able to replicate
Andrew Curran@AndrewCurran_
Striking image from the new Anthropic labor market impact report.
English
English
@ycombinator @Heypocket @AkshayNarisetti @gabrieldymowski Amazing product, used multiple times!
10x improvement of offline meetings productivity!
English
Pocket (@heypocket) is your notetaker for real world meetings.
In the last 5 months, the team has delivered over 30k units with a $27M annualized run rate, growing 50% month over month.
Congrats on the launch, @AkshayNarisetti and @gabrieldymowski!
ycombinator.com/launches/PaX-p…
English
i hate the whole "iT's NeVeR bEeN eAsIeR tO rAiSe cApItAl" narrative because that seems to apply exclusively to grifters leveraging the AI hype, whereas:
Agriculture innovation -> all-time low
Diagnostics (hardware) -> all-time low
Biomanufacturing -> all-time low investments
your grifts are cannibalizing all available capital, and we won't have anything to eat in eight years; but sure, go raise capital for AI agentic-driven marketing automation for CRMs
English
@prywatnik Czy tylko dlatego że praca jest niereprodukowalna?
Polski
Żaden LLM czy AI nie odtworzyłby mojej rozprawy doktorskiej z informatyki, obronionej ponad 10 lat temu ;-)
Polski
Greg Preibisch, MD retweetledi
sitting downstream of the frontier labs on the value chain is a dangerous place to be imo (even in the life sciences)
specialized verticals would seem to be more insulated, but look at how many connections anthropic is making into downstream specialized knowledge applications in life sciences (quoted list below)
as we've seen in codex, claude code, claude code security, and more - frontier labs are hungry to make your vertical platform a feature on theirs
from a frontier lab perspective, the economics are driving here - as open source LLMs are commoditizing the knowledge api layer, application layer is naturally where you have to go to capture value and justify valuation
this becomes a race of how fast downstream players can build distribution vs how fast life science teams at anthropic, openai, deepmind etc can build features / connections
and since life sciences (and healthcare) is generally a b2b sell (industry R&D researchers don't do IP related work outside of work digital environment), virality is (generally speaking) out of the question
good luck to all in the arena 🙏
Healthcare AI Guy@HealthcareAIGuy
Anthropic’s healthcare & life sciences strategy map 1. Investments • Phylo – AI-native biotech platform • Heidi – AI ambient scribe 2. Partnerships • HealthEx – Health data exchange • Genmab – Antibody therapeutics & R&D • BioRender – Scientific visualization • 10x Genomics – Single-cell biology • Benchling – Biotech R&D cloud infra • Allen Institute – Biomedical research • HHMI – Life science research
English