Always wanted to see Rinzel as the 2nd line center.

I found this to be just as evident with reading. I spent some time having AI summarize new papers. I found it did a decent job! Only a month later I had retained none of the information of those papers. I had spent no time milling with the ideas, figures, or experiments. Useless.

When the filter is removed (change to non-selective environment), a higher dimensionality of pleiotropic effects becomes apparent.

The study finds no evidence for this hypothesis, and instead supports the "pleiotropic shift" model.

My understanding is this is correct. In their pleiotropic expansion model, the mutations all have a broad inherent pleiotropy, but the selective environment acts as a filter, allowing only a low-dimensional subset to be relevant for fitness in that environment.

The idea would be a sphere is perfectly captured by the l=0 harmonic, while less spherical shapes will required increasingly higher order harmonics to accurately describe. Maybe more resolution/information than surface area vs volume, but less intuitive to interpret for readers.

This might be something that decomposing the shape into its spherical harmonics could do quite well.... It's kind of like how a Fourier transform breaks a complex signal down into sines and cosines, but it's breaking a 3D shape down into orthogonal shapes on a sphere.

Really cool! Congratulations!

Ba6, c4, Rf3, b3, Rf8

Black looks really good but after c4 gumming up the nice mate, I'd find a way to lose this :)

Thanks Willem!

For what it's worth, chrome and firefox both have extensions to high google AI, which is typically what I have going on.

Google AI is horrible. You alter the search even a little bit and the answer is dramatically different. I wish duckduckgo was a slightly better search engine.

As a final note, I just want to drop a shout out to Anh (not on bsky) who helped make this possible after I graduated. She did some important experiments during the revision process and she's a truly spectacular scientist.

As a result, collateral profiles are highly dynamic/variable through evolutionary time and we need a stochastic control model to optimize treatment. To do with we adapted our 2019 Markov Decision Process model to include temporally varying state profiles!

We see that adaptation to antibiotics results in near uniform collateral resistance, further adaptation results in both collateral sensitivity and resistance. This is very similar to recent work from @grantkinsler.bsky.social @petrovadmitri.bsky.social in their yeast two-step paper (also in PLOS B)!

Joe Scally 4 times?