Alex Rubinsteyn

I dunno, but it’s the kind of hubris that kicked our butts in Hammer Lab

“Suppose it takes a year for the clinical trials to work before you’re getting revenue from drugs”

LOL

Making myself finish the Dario/Dwarkesh podcast bc it feels important to hear a main character interview but…it’s surprisingly thin. Maybe I’m already too familiar with his essays and he’s otherwise holding cards close?

open-source the means of production

Starting to see cracks emerging all over the traditional models of drug development, clinical trials, treatment/delivery, “reimbursement”.

It’s chaotic but generally patient agency seems to be increasing alongside rate of medical progress.

Like emerging from a long dark tunnel

Emotionally, going from CS to biomed is rough.

Everything is slower, hierarchical, bureaucratized, entrained in Byzantine incentive structures, pay is bad for important roles bc of a tax on passion, public comms obfuscates behind concern for patients, &c

A small polymerase ribozyme that can synthesize itself and its complementary strand The emergence of a chemical system capable of self-replication and evolution is a critical event in the origin of life. RNA polymerase ribozymes can replicate RNA, but their large size and structural ...

word of the day: eutectic ice

www.science.org/doi/10.1126/...

pubmed.ncbi.nlm.nih.gov/12448990/

Trial was revised many times, some of which look deeply confounding

eg v6 to v7 changed trial from observational to interventional and from immunotherapy alone to immunochemotherapy

@quaidmorris.bsky.social want to collab? We have a long read RNA variant extraction tool that gives me all differences from reference in long read single cell RNA seq.

I want to pull this apart into {somatic, germline, noise} by EM (alternating with cell clustering). Does this already exist?

Luckily/unluckily the trial looks very unsound

If you want any kind of say in what kinds of social rules and norms get put around a new technology you have to at least understand what it is.

Like, cars had a hideously distorting effect on the urban fabric of most countries but there’s no way you to mitigate that outcome by denying they go fast

Like, agriculture was like this (read Against the Grain), creating both oppressive social hierarchy and ecological destruction. Pastoralism / animal husbandry was even more ecologically destructive. Mining, metallurgy, industrial manufacturing, fossil fuels, plastics, &c

Yes!

We live atop a big stack of very destructive and society-rearranging technologies which are also inseparable from our material culture and standards of comfort. I guess a new one can be hard to think clearly about but there’s so much precedent.

I don’t understand why it’s difficult though.

Like, you can just try out all the tools and see what they’re good at, then think a little bit about the consequences

The social direction will be hyper-oligarchy -because- the tools are useful. If someone can’t see that it’s very weird

Moralizing AI to the point of refusing to understand it seems like such a political miss.

Biggest realignment of capital / labor / social norms & institutions in a long time…

I don’t think that’s quite right though, my Max & Pro subscriptions are squarely opex, just not wages. It’s turning labor into SaaS opex (except for the handful of companies serving the models, who have to spend a Manhattan project on capex)

I think this will be a big unlock for us.

Working on: EM alternating estimation of cluster membership vs variant category assignment

And, not just recycling cliches or circularly referencing a pile of claims based on that one time someone tried 4o and it was somehow immoral

Like, who’s got Codex running in six tabs but is also talking about the social trend lines or even organizing towards the coming challenges?

Is there an AI savvy corner of the lefty political milieu?

Like, obviously this stuff is going to accelerate oligarch-ification, is there anyone publicly thinking about social consequences who has a grounded understanding of the frontier and trajectory?

I still have trouble with Anthropic models hallucinating references sometimes, which I haven’t seen in like a year from the o3 / ChatGPT Pro line of models

I left xAI today.

I will never forget the friends I made in the extracorporeal deputrification chamber.

A brief reset and then onwards! 2026 will be the year we finally convert all living biomass into a perfect sphere. 🚀🚀🚀

Crushing

For the purpose I want (skipping DNAseq) it won’t be sufficient but yeah conceptually you could still a subset and in a high TMB cancer use that to confidently split out tumor cells

real

You can get germline + somatic together from Exacto (my lab is making it), need to make something on top to deconvolve tumor from normal

Interesting, which model is this?

(Expectation Maximization)

No! A magic other thing

Just realized you don’t need DNA to find almost all expressed somatic variants if you have long read scRNA

Codex kicking off large training jobs on Modal and I feel like I'm ascending into a Sci-Fi hallucination