Nathan Insel

Nice!

Or is she transcendendate

If the purpose of analysis is to test hypotheses, then model should come first.
If the purpose is to make new observations (to inform models), then after.
But neuro tends to be pretty bad at separating the two, and stats or model fits sometimes become tautologies

There are many pitfalls and challenges. Nachum discusses this in the book, and one chapter is about technological developments that make it more feasible (this is also why @wheelerlab.org and I put some years into developing DIPLOMAT). But he also describes so many ways it has paid off

I realized it's because I cared about how brains deal with layers of uncertainty by forming and updating many levels of expectations. And a good way to get uncertainty, along with the behavioral agency to address it, is letting animals behave freely in dynamic, uncontrolled environments.

Personal note: when I started my own lab I had a small crisis: why study extremely complex brains when we still don't understand how simpler brains do the same things?

The back-and-forth between "control all variables" and "observe systems as they naturally occur" is probably as old as science. Nachum explains so many reasons, with so many elegant examples, why systems neuroscience needs a diversity and balance of approaches.

One of the best books I read in 2025 was Nachum Ulanovsky's _Natural Neuroscience_. If my copy looks a little worn, it's because I was intent on reading it in natural habitats.

Generative AI creates synthetic regulatory DNA sequences for precision gene control - Nature Genetics We developed DNA-Diffusion, a generative artificial intelligence (AI) method that creates synthetic regulatory elements showing enhanced activity. Multiple synthetic elements demonstrated superior cel...

www.nature.com/articles/s41...

A non-scientist friend saw your name in this article and used your quote to counter something I had previously said. She says thank you.

Great culture can save lives. Literally.

Amazing letter in today’s @thetimes.com about Tom Stoppard

On the Nobel Prize winner, Camillo Golgi (from Cani and Mazzarello, 2016)

Seller's profile pic:

Amazing! Congratulations!

There are certainly other good solutions to multi-animal tracking that DIPLOMAT does not yet implement--like finding visual differences between animals. All code is open source (github.com/TravisWheele...) and these features could be easily incoporated or used alongside the algorithms packaged here.

How it works: instead of finding body parts within frames and then knitting these across frames, DIPLOMAT applies a hidden Markov model (Viterbi algorithm) that traces body parts through a video. We use the same movement and skeletal information as other tools, but as HMM transition probabilities.

Perhaps most useful is that DIPLOMAT is designed to make manual edits much more efficient: multiple body parts can be corrected with a single click and, once corrected, re-traced across frames.

With DIPLOMAT, you can still use models trained in DeepLabCut or SLEAP!

But the algorithms also improve tracking (fewer body-swaps, better recall and precision)--even when body parts are occluded.

DIPLOMAT: multi-animal tracking with efficient manual editing Recent advances in computer vision have enabled the development of automated animal behavior observation tools. Several software packages currently exist for concurrently tracking pose in multiple ani...

New preprint:
There are some decent tools for multi-animal tracking, but it can still be difficult to track interacting animals without mixing them up. Isaac Robinson (with @wheelerlab.org and others) have developed a software solution that we are now using full time: doi.org/10.1101/2025...

More biomed in Canada

"Emotion" vs "affective state" sounds like a fun example of how psychological constructs are clustered hierarchically. Would be interested if anyone takes this up in an articulate and compelling way.

Not an expert, but I take this issue up in my intro neuro course. We walk through ~8 definitions and examine where they fail. We then put these together into our own, including "usual" causes (perceived valence) and "typical" consequences (visceral, also facial/vocal expressions).

New strategy for grant and paper rejections

Lemborexant ameliorates tau-mediated sleep loss and neurodegeneration in males in a mouse model of tauopathy - Nature Neuroscience Parhizkar et al. show that lemborexant, an orexin receptor antagonist, protects against neurodegeneration in male tau transgenic mice by preventing tau protein build-up and inflammation, highlighting ...

Notable paper on sleep aid lemborexant (putting aside caveats of mouse models of Alzheimer's). I have collected 2 1/2 years of data on myself and can confirm even small (2 mg) doses add sleep hours.
www.nature.com/articles/s41...

Thanks for this!
On writing: I remember Tulving saying that he admired and was inspired by Hartline's paper on lateral inhibition for its writing. From all the written history about Tulving and his influences, this isn't something one might easily guess.

Oh so now we are in a parasocial friendship?

Cool!

Didn't catch that, thank you.

Will add "toothpaste related crime" to the landscape of imagined futures

Earnest question: can't kids get their fluoride by regularly eating a small amount of their toothpaste?

To borrow from Alan Moore "Uglier than death backin' outta the outhouse readin' mad magazine and crazy as a football bat."