Thomas Lau

@sethbannon @baym en.wikipedia.org/wiki/Combinato…

@baym What do you mean?

If only there were some way to do chemistry combinatorially

poetry sent by a friend. poignant

I wrote up some thoughts on the automation of lab work, in particular how it relates to how people will work in the lab. In short, it will deliver a lot of value for assays run at scale, but there is a long tail of experiments where humans are essential. andrewcarroll.github.io/2026/02/09/for…

A typical project for an incoming graduate student might involve 1–2 weeks of planning and 2–5 years of execution, [yet] the problem you choose will influence the impact of your work just as much as the quality of your execution. – Michael Fischbach buzzsprout.com/1744020/episod… @mfgrp

Excited to present the first major work after starting our lab at Stanford and the Arc this year: CRISPR-All, a unified genetic perturbation language for programming any major type of genetic perturbation simultaneously, in any combination, at genome scale, in human cells.

Has anyone even asked the cells if they wish to be perturbed

@luca_mchtt @pochekailov Especially when you're going to spend $1K+ on sequencing, it feels risky to try to save $20 on SPRI beads. I've stuck with the Zymo beads so far but would be open to trying new ones if there was an easy way to get a free sample on the website.

@pochekailov Hi Sergii, we've ordered your racks and they are so much better compared to commercial ones, for both cost and design! We never tried the beads, I guess since it's for library prep related stuff and we have set workflows and reagents for that, it always feels risky to change

I have been selling magnetic racks and magnetic beads for several years now. Magnetic racks are about 65% of sales. Magnetic beads for DNA purification are only 9%. I assume that you use the racks mainly for DNA purification. Am I wrong? If not, is there something wrong with my beads? I would expect much more magnetic beads sales. I would greatly appreciate to learn why you would not choose my beads. Price too high? Or too low? Too good to be true? You tested the sample of beads with NEB ladder and it didn't work? Something else?

@aaronmring @AdrianoAguzzi @pochekailov makes very affordable magnetic separators. I’m a big fan of these. sergilabsupplies.com/products/96-we…

I'm now the proud owner of a bespoke, handcrafted 96 well magnetic separator designed by my Twitter BFF @AdrianoAguzzi. Thanks a million!!!

@JessicaSacher @OpenAI Same! I’ve moved over to Gemini.

Cool now getting blocked from basic questions about finding other biology researchers who work on bacteriophages @OpenAI please fix this so scientists can use ChatGPT again Are other scientists getting blocked like this?

@KerenGu For me, o3 has recently gone from "great for brainstorming ideas" to rejecting half of my prompts when it comes to biology research. @KerenGu can we have a verified user system for biological capability? (TSA Precheck?)

We’ve activated our strongest safeguards for ChatGPT Agent. It’s the first model we’ve classified as High capability in biology & chemistry under our Preparedness Framework. Here’s why that matters–and what we’re doing to keep it safe. 🧵

@plasmidsaurus Does ZeroPrep support EndA+ strains now?

Wait, you’re still doing minipreps? 💀 Use ZeroPrep to go straight from colony to whole plasmid sequence, overnight. Now supporting low-copy plasmids — so you can save time, reduce labor, and tackle constructs. Start sequencing our improved service now: plasmidsaurus.com/products/zerop…

bias-variance tradeoff

All I said was Bayesian statistics are inferior to Frequentist statistics.

Trust me, I am a Bayesian.

It should be noted that, in combinatory logic, the semantics of the Y-combinator is to endlessly repackage a function (f[y[f]], f[f[y[f]]], ...) ad nauseam, making no other modifications.

@minchoi The first protein folding example is satire, o1 cannot perform protein folding.

OpenAI o1 is wild. It's only in "Preview" without vision and people are already doing incredible things and enhancing their workflow. 10 examples:

@miniapeur John Duchi’s course at Stanford STATS 311: Information Theory and Statistics has a great series of lecture notes. web.stanford.edu/class/stats311…

What is the deep connection between information theory and probabilities?