Surely people training on Wikipedia would just download the corpus, which is trivial to do, and analyze their local copy, which is orders of magnitude faster than crawling the website?

Yet one more reason we cannot allow LLMs to serve as epistemic grounding is that we cannot triangulate among them the way you can among reasonable independent sources. They bullshit in the same way and end up agreeing with one other about things that are completely false.

The manuscript is out now--take a look! www.nature.com/articles/s41...
We’d love feedback! As always, please reach out if you’re interested in getting more info or reagents. 10/10

We have transcriptomics data describing the on- and off-target effects of different AID system implementations. We also generated several new TIR1 lines we think will be useful to others, including a new “whole-body” line with pan-somatic and pan-germline expression. 9/10

One of the most exciting outcomes was that we figured out how to control two different proteins independently in the same living animal, a “dual channel” AID system that opens the door to cool new experimental designs. 8/10

Along the way, we ran into several surprises including some unexpected (and in some cases unfortunate) features of the popular AID2 system...it's worth taking a look at, for anyone using the AID2 system. 7/10

So we took a close look at existing AID implementations for use in quantitative degradation experiments. We ran careful dose–response experiments, and found it is possible to control the rate of an animal’s aging by quantitatively degrading components of IGF signaling. 6/10

Most people use AID in a binary way: adding lots of auxin to drive protein levels as low as possible. However, biology—especially aging and physiology—is rarely just “on” or “off.” What my group really wanted was a way to tune protein levels in a controlled, quantitative way.5/10

In 2009, the yeast biologists Kanemaki et al realized that they could take TIR1 out of plants and put it into other organisms--degrading any protein modified to include an AID tag. In 2014, Abby @adernburg.bsky.social, Jordan @gotworms.bsky.social, and others adapted it for C. elegans 🐛. 4/10

Our approach builds on the Auxin-Inducible Degron (AID) system, a technology derived from plant biology, where a protein called TIR1 triggers the degradation of any protein containing a short target peptide sequence—but only when activated by a specific hormone called Auxin. 3/10

Our lab studies the molecular genetics of aging--we want to understand why organisms grow old and how to engineer them to age better. 🛠️🧬💀That said, the technology I’m describing here should be useful well beyond aging, for anyone doing quantitative biology in eukaryotes. 2/10