almost everyone on HN is interested in biology and especially biohacking but does not actually realise they have none of the necessary expertise to judge claims coming from that field; the level of discourse on these topics is sometimes worse than what I spot on reddit

Pish Posh! My understanding of C pointers makes me one of the smartest people in the world! Why else would I have cleared the hurdle at $COMPANY during a tremendous bull market!

Clearly I can learn any discipline by mucking about for a few afternoons. What is a billion years of natural selection to my towering intellect?

Lots of people said this about SpaceX and Tesla. Besides, our towering intellect did destroy millions of years of Small Pox evolution.

Turns out some people are smart. Other people who do the same thing are stupid but it isn't a marker.

To be lovingly snarky: HN in aggregate is incredibly skeptical of academic research, claiming reproducibility crises everywhere, misalignment of incentives, brutal working environments whose results can't be trusted, only in mice, etc.

Unless it either confirms their existing beliefs, or is so technical that they don't understand it. Then the research must be ok :)

Yes, when a topic you're intimately familiar with comes up, it's unbearable reading the comments.

Worst part is when you try to set the record straight and wind up with -5 votes and a dead comment within minutes.

I concur; the graphs didn't show much difference to me (but I was on my phone so I couldn't be certain :-) The study seems impressive from the title, but reading it not so much. The Alzheimer's patients in the study are older, more male, and less educated than the controls, so it seems quite likely that they are eating worse which would explain the different microbiota. This could mess up the mice in multiple ways. (Seriously, the mice that receive Alzheimer's microbiota could have a stomach ache and do slightly worse on the tests.)

As for why there is so much interest on HN, I've noticed that Alzheimer's is one of the topics that is overly popular on HN, along with category theory, Lisp, and problems with Tesla, to name a few.

Not really weighing in one way or another, but I do not see scatter plots in the paper other than one PCA plot (which was not LDA... not sure why they chose PCA, so no strong reason to expect point separation there, though I do not know the diversity metric they are using so feel free to ignore my commentary).

Violin plots always look like they overlap. I think seaborn's split=True should be pushed a bit harder.

Figure 3 is the one showing the cognitive results.

This field is littered with studies like this, measuring hundreds of different outcomes and then presenting the statistical noise as positive findings.