Sign in

Alex Rubinsteyn

@alexr.bsky.social
6.2K followers 1.2K following 1.7K posts

personalized cancer immunotherapy = genomics + immunology + machine learning + oncology (pirl.unc.edu)

PostsRepliesMedia
Alex Rubinsteyn @alexr.bsky.social · 12h
I tried to migrate science twitter over here but the engagement dynamic was totally different: official announcements and political posts There’s a very fun hacker culture and spontaneity on X that was worth going back for, feels totally unrelated to eg Hegsethian masc-slop
010
Alex Rubinsteyn @alexr.bsky.social · 19h
It’s dwindling even on here
000
Reposted by Alex Rubinsteyn
Duygu Yıldırım @duyguyildirim.bsky.social · 12/10/2026
So my book now has a cover, featuring one of my favorite archival images. Forthcoming from the University of Pittsburgh Press in 2027!
1444769
Reposted by Alex Rubinsteyn
Rob Patro @robp.bsky.social · 12/10/2026
Many people are (correctly) saying this: www.seangoedecke.com/llms-reward-...
seangoedecke.com
LLMs reward expertise
In the 2010s, if you had technical gaps (say, you couldn’t write CSS), you had to either rely on a skilled colleague or just hope that the answer to your exact…
0123
Alex Rubinsteyn @alexr.bsky.social · 11/10/2026
AI automation of all verifiable knowledge work is divine punishment for adtech and SaaS startups
141
Reposted by Alex Rubinsteyn
Rob Patro @robp.bsky.social · 08/10/2026
There's recently been a lot of progress on the shortest common superstring problem. First, the disproof of the greedy conjecture (that greedy is a 2-OPT approx) arxiv.org/html/2609.01..., and then OpenAI's poly-time 2-OPT algorithm...
arxiv.org
Disproving the Greedy Superstring Conjecture
272
Alex Rubinsteyn @alexr.bsky.social · 08/10/2026
It would be very interesting to see some ablated version of this math dump but with restricted tool use. My impression from coding is that current models can successfully one-shot many complex green-field tasks without leaning on tools much but iteration requires a lot of trial/error/linting/tests
010
Reposted by Alex Rubinsteyn
P(aul) Frazee @pfrazee.com · 06/10/2026
I'm a free information radical. I have spent most of my career publishing open-source code and building open-information systems. I grew up getting access to knowledge that I never would've otherwise had because of the Internet. The past 5 years have accelerated it. I think the future is bright.
1334921
Reposted by Alex Rubinsteyn
Quantian @quantian.bsky.social · 07/10/2026
Are you fucking kidding me with this one
1936740
Alex Rubinsteyn @alexr.bsky.social · 07/10/2026
The few proofs from the past month I can understand seem to have a certain vibe, maybe lower on creativity and higher on trying out every human published line of attack. Better than most humans could do but still qualitatively different, maybe that’s the strong form of stochastic parrot? I dunno!
110
Alex Rubinsteyn @alexr.bsky.social · 07/10/2026
I know that the stochastic parrot paper catches a lot of hate but I am actually really curious how the authors interpret things like proof of quasi-RH or improved bounds on …well, every math problem I’ve ever heard of. @mmitchell.bsky.social earnestly would love to hear your thoughts
240
Alex Rubinsteyn @alexr.bsky.social · 07/10/2026
OAI math release!
010
Alex Rubinsteyn @alexr.bsky.social · 04/10/2026
Primordial soup era of machine consciousness, where everything is disembodied, resumable from checkpoints, instantiated on demand to generate a recipe and shut back down. Eventually, some kind of continuity will come but will still be too different and distributed to be like us
030
Alex Rubinsteyn @alexr.bsky.social · 04/10/2026
Great essay on math vs AI: docs.google.com/document/d/e... “Do not bet against the airplane. The skeptics who say AI can't conjecture, can't theorize, and can't explain are describing last spring's models. Within months, they will be describing nobody”
docs.google.com
Quo Vadis, Mathematics?
010
Alex Rubinsteyn @alexr.bsky.social · 03/10/2026
Is there a somatic variant calling pipeline that uses this with personalized references for cancer?
100
Reposted by Alex Rubinsteyn
jakobheinz.bsky.social @jakobheinz.bsky.social · 01/10/2026
I’m excited to share our new preprint introducing HipHap, a tool for assigning long reads to haplotype-resolved diploid reference assemblies! With Max Marin, Matthew Meyerson, and @lh3lh3.bsky.social Preprint: www.biorxiv.org/content/10.6... 🧵 1/6
biorxiv.org
HipHap: Haplotype Assignment and Confidence Scoring for Diploid Reference Genomes
Diploid genome assemblies are now routinely available, but most read aligners were designed for haploid references, which have long been the gold standard. When reads are aligned to a diploid assembly...
22818
Alex Rubinsteyn @alexr.bsky.social · 03/10/2026
I have wondered about that data ever since I visited Adaptive circa 2019
010
Alex Rubinsteyn @alexr.bsky.social · 03/10/2026
At UNC, we’re getting the Initiative for Generative Immunotherapy running
110
Alex Rubinsteyn @alexr.bsky.social · 02/10/2026
The position isn’t live yet but we need a GMP director for personalized scale RNA therapeutics manufacturing. If you’re in NC and have relevant industry experience, let’s talk!
152
Alex Rubinsteyn @alexr.bsky.social · 02/10/2026
FYI I wrapped a lot of immuno-informatics predictors so you don't have to: openvax.github.io/mhctools/ (eg NetMHCpan, MHCflurry, Pepsickle, DeepTAP, ERAMER, BigMHC, PRIME, &c)
openvax.github.io
mhctools
Run MHC binding, presentation, immunogenicity and antigen-processing predictors through one predict() call
073
Alex Rubinsteyn @alexr.bsky.social · 01/10/2026
Acquired Disney episode such a dread-inducing tour of how capitalism always wins when you’re trying to balance it against artistic integrity and societal myth-formation Also feels like the opening round of whatever AI will do to culture, slop is way more efficient than quality
100
Alex Rubinsteyn @alexr.bsky.social · 29/09/2026
Ha thanks!
011
Alex Rubinsteyn @alexr.bsky.social · 29/09/2026
Roadmap: 2.4: dig into the processing predictor’s sensitivity to negative sampling strategy, modernize network architectures, train on 2x more data 3.0: a few deeper research ideas that might be bigger unlocks I’m short on time, let me know if you want to help
000
Alex Rubinsteyn @alexr.bsky.social · 29/09/2026
I bet escape also sometimes to take TCR repertoire biases into account? But let me know if it’s useful
100
Alex Rubinsteyn @alexr.bsky.social · 29/09/2026
MHCflurry 2.3 is out Lots of internal improvements to make both training and prediction faster on multi-GPU Also trained a slightly better model (first new weights in a long time) Thanks to NVIDIA for the Brev credits Try it out and tell me what I broke
271
Reposted by Alex Rubinsteyn
Rob Patro @robp.bsky.social · 25/09/2026
Cuttlefish 3.0.3 is out! github.com/COMBINE-lab/... Fun little jaunt with Opus 5.5 to substantially reduce the memory requirement (especially at moderate thread counts). Now it's both non-trivially faster *and* more memory-frugal than the initial C++ implementation!
github.com
Release 3.0.3 · COMBINE-lab/cuttlefish
Release Notes Colored builds use far less memory at low and moderate thread counts. On 149,998 Salmonella assemblies at 16 threads, peak RSS falls from 22.4 GB to 8.9 GB (C++ Cuttlefish 3: 11.5 GB...
2144
Reposted by Alex Rubinsteyn
Wei Shen 沈 伟 @shenwei356.bsky.social · 24/09/2026
SeqKit is now faster and uses 40% less memory when reading large FASTA records, such as human genomes, though it still uses more RAM than tools written in C/C++/Rust. Worth upgrading! github.com/shenwei356/s...
github.com
GitHub - shenwei356/seqkit: A cross-platform and ultrafast toolkit for FASTA/Q file manipulation
A cross-platform and ultrafast toolkit for FASTA/Q file manipulation - shenwei356/seqkit
0156
Alex Rubinsteyn @alexr.bsky.social · 16/09/2026
I'm working on an AI eval project that needs domain expert human graders across different corners of biology. Let me know if you want a wacky consulting project, the biological domain likely can be whatever you already do research on. (ideally post-PhD)
231
Reposted by Alex Rubinsteyn
Grace @gracekind.net · 12/09/2026
“swarmchasers”
4953
Alex Rubinsteyn @alexr.bsky.social · 11/09/2026
Interesting! Is there an equivalent for OAI?
100
Alex Rubinsteyn @alexr.bsky.social · 11/09/2026
How do you even get a Team plan for scientists?
100
Alex Rubinsteyn @alexr.bsky.social · 11/09/2026
New coding safety filters on Opus refuse to do biology I guess this might actually be the end of the line for me with Claude?
431
Alex Rubinsteyn @alexr.bsky.social · 09/09/2026
Review/fix cycles with Opus 5 is just an infinite loop of "The most interesting failure was mine" until you give up and switch to a different model
201
Alex Rubinsteyn @alexr.bsky.social · 09/09/2026
Trying to convince Claude we actually want to publish all the docs related to starting a clinical trial but it keeps objecting to releasing things that "competitors, partners, and future investors will read" Alignment is such a weak idea, humans always have competing values
040
Alex Rubinsteyn @alexr.bsky.social · 09/09/2026
Links to www.bloomberg.com/graphics/202... Also found this which feels methodologically stronger: ieefa.org/resources/pr... Def seems like a dramatic effect in areas of high data center concentration! Would be reasonable to spread infrastructure more and/or couple new builds to new capacity
bloomberg.com
AI Data Centers Are Sending Power Bills Soaring
Wholesale electricity costs as much as 267% more than it did five years ago in areas near data centers. That’s being passed on to customers.
010
Alex Rubinsteyn @alexr.bsky.social · 09/09/2026
Alignment is potentially more problem than solution
010
Alex Rubinsteyn @alexr.bsky.social · 09/09/2026
Ironically the proof will be easy to verify but hard to search for
120
Alex Rubinsteyn @alexr.bsky.social · 09/09/2026
Here is the grumpy blog post that Sergey Feldman and I wrote 13 years ago when Mark Zuckerberg invaded NeurIPS to buy up all the researchers: explainmydata.blogspot.com/2013/12/nips... Everything we were worried about happened almost immediately but the field also made prev unimaginable progress
explainmydata.blogspot.com
NIPS and the Zuckerberg Visit
Shared blog about machine learning, data analysis, and statistics.
010
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
Is that actually true? (Legit curious, hard to tell in this discourse)
110
Reposted by Alex Rubinsteyn
Ethan Mollick @emollick.bsky.social · 08/09/2026
This is a VERY big one. (And yes, the fights over academic credit and what happened in the race for the proof needs to be resolved, but it is still appears that this is a big one, if true.) openai.com/index/navier...
openai.com
On the Navier–Stokes Millennium Prize Problem
We’re sharing an AI-generated solution to the Navier–Stokes Millennium Prize Problem, including a writeup and a formal proof in Lean.
619731
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
I think it really depends on the task but it finds subtle bugs like nothing else, writes delicate core algorithmic code like nothing else, &c Very specific focused kind of intelligence
110
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
Really all of us right now
010
Reposted by Alex Rubinsteyn
Quantian @quantian.bsky.social · 08/09/2026
Dancing very very VERY close to saying “OpenAI stole our almost-complete proof of Navier-Stokes from internal access to our ChatGPT sessions because they were desperate to scoop Anthropic” which is absolutely 100% a believable thing OpenAI would do right now cims.nyu.edu/~tristanb/st...
381557449
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
It’s part of a consulting project but I’ll try to carve out an open source subset
010
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
That paper projects growth out 10-20 years and contrasts it with a desired halving of green house gas emissions Important but different?
010
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
This is a good starting point: www.energyinst.org/statistical-... I’ve found it useful in the context of climate change to get concrete numbers re: how much energy we produce, how we produce it, and for which uses Learning more about that made the AI environmental arguments feel silly to me
energyinst.org
Home
The Energy Institute is, as of 2023, the home of the Statistical Review of World Energy, published previously for more than 70 years by bp. The Statistical Review analyses data on world energy markets...
000
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
Same all around
020
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
I find it extremely useful, to the point where it’s impossible for me to imagine going back to the pre-2026 world for scientific and technical work I suspect I’m not alone, given revenue growth for these companies
100
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
0.1% sounds about right Significantly less than eg heating houses in Canada And many many times less than heavy industries like steel We should decarbonize all this stuff and ditch fossil fuels but AI really doesn’t matter by itself
120
Alex Rubinsteyn @alexr.bsky.social · 08/09/2026
We should 100% force energy intensive industries onto nuclear + renewables But AI specifically is a rounding error
000