@lpachter.bsky.social, a Good Human Being who is also a scientist, put it very, very well in a blogpost

There are many moments when you sit back from your nth Tchaikovsky novel to think "actually this guy simply prints incredible lines"

Assembling the shiv exodia is one of the best feelings

I see you've committed the grave crime of *checks notes* being a woman on the internet.

Maybe keywords in the title triggered some kind of additional checks behind the scenes given the climate around billionaires right now

Favorite example of this was Smith-Waterman in Geology. I believe the anecdote was that they were incidentally walking by a local Geo meeting and saw poster presentations describing the problem, and realized that it was just another application of sequence alignment. The story is all hearsay though

If you're trying to cluster one underlying truth with two different methods, as long as both methods _can_ reach the same defensible answer, does it matter? Correct me if I'm wrong but you're describing how methods can be wrong, which is separate from different methods giving different answers

As a not-scRNA person, is this something that happens when one uses defaults and calls it a day? Can parameter tuning of clustering bring the results in line with the expected/observed biology or are they simply wholly incompatible? As I understand it, scRNAseq analysis is quite iterative

KSP2's history is a constant repeat of "and then it got worse"

Fighting over 2nd place while data.table looks on from the top of the podium?

Considering how many prescient things they did predict, it's not a bad track record though. It makes sense they'd expect brainwaves to be decodable far more easily than in reality

Terminals are actually terminal WHAT