MillironX

I very much have had the same experience as the author with research libraries. I have learned so many life-changing techniques and derivations from books that were two or three down from the book that I was looking for. When I became a research associate, I would often have my librarians ship the books to the branch closest to my office, and I stopped browsing the shelves myself. It wasn't until I went to the main library that I checked out a book several over from the one I had found in the catalog that I had a major breakthrough. Now that same library has collapsible shelving that makes browsing feel like a chore, and it makes me think we lost something valuable.

That said, I had high hopes for the catalog that the author had put together, but for my technical fields, I found it lacking. Trying to find books that would be close to van Soest's Nutritional Ecology of the Ruminant or van der Werf's Animal Breeding: Use of New Technologies, yields a whole bunch of pop-sci titles. The joy of finding these titles at the library in the past was how remarkably understandable and engaging they were while remaining scientifically rigorous. I immediately feel that the pop-sci titles that I was recommended are going to try to sell me something or some idea, rather than letting me explore ideas myself. I guess that means I need to go back to browsing library shelves myself.

We covered yew extensively in toxicology class in vet school, but I didn't know about any animals that eat the berries. My favorite fact about yew is that the Iowa State Lloyd Veterinary Center is named after a toxicologist, yet has yew planted for decoration all around the building.

Of the communities mentioned that I recognized, most were retirement-aged communities. In that case, wouldn't age demographics alone explain the population differences?

My personal gripe with Ente is that they seem to value making new half-baked products over making their current products feature-complete. Case-in-point: RAW photos are still not supported in Ente Photos [1]. Locker and now Ensu got priority over a key feature in making the Photos app a photos app.

[1]: https://github.com/ente-io/ente/discussions/625

My college has replaced textbooks with internet resources and "current academic literature," but I'm shocked at the number of people who can't interpret a scientific article, nor properly search a database like PubMed or CABI. It can't be for lack of training: we have entire semester-long courses dedicated to research and library use. Maybe the distractions of text messages and TikTok are to blame like the article suggests.

In the animal world, the genomic vaccine revolution has led to two new types of vaccine being produced.

1. The Autogenous vaccine, where a veterinarian submits swabs of offending organisms to a lab that cultures and produces a vaccine specific to that strain to sell back to the clinic. Genomics play a role in determining if the currently circulating strains match the culture that the vaccine is based on.

2. The prescription platform vaccine, where genomic sequences are used to create a custom antigen that is produced in a microbial factory.

The efficacy of these types of vaccines is hotly debated because they do not have to go through clinical trial before distribution. The development of "commercial" (i.e. clinical trial-burdened) vaccines appears to have slowed based on my observations. While I think some of these new vaccines show promise, many autogenous and prescription platform vaccines are just more expensive versions of vaccines already available, or they package organisms that are suspected commensals and market it as something that no one else has done.

It's hard for me to call today a "golden age of vaccine development" in that light. I hope that the article is only taking human vaccines into account and that there is appropriate efficacy data for every vaccine shown in the last chart.

> A key assumption in my argument is that the marginal benefit of spending more time on an exam is always nonnegative

That is a very bold assumption, and I doubt the data supports it.

I couldn't find much published research on the correlation, just this one paper: https://journals.sagepub.com/doi/pdf/10.2304/plat.2009.8.2.5.... It shows no correlation between time on exams and final score, but its methodology is questionable.

The first argument actually leans in favor of LaTeX or Typst as a better replacement for Docx.

A LaTeX or Typst document can contain both the content and formatting together within the same file. This isn't idiomatic for either language, and my experience is that this is more common for Typst than LaTeX, but both can do so. All of those formatting rules like small caps, table widths, margins, page numbering, etc.? Those can be rigidly defined in either LaTeX or Typst and are better guarded aginst accidental formatting rules breaches from double click, copy/paste, or table cell insertion than in Word.

I'm more sympathetic to the network effect argument. It's hard to envision a reasonable redline system compatible with both Docx and LaTeX/Typst.

> I don't think Julia really solves any problems that aren't already solved by Python.

But isn't the whole point of this article that Matlab is more readable than Python (i.e. solves the readability problem)? The Matlab and Julia code for the provided example are equivalent[1]: which means Julia has more readable math than Python.

[1]: Technically, the article's code will not work in Julia because Julia gives semantic meaning to commas in brackets, while Matlab does not. It is perfectly valid to use spaces as separators in Matlab, meaning that the following Julia code is also valid Matlab which is equivalent to the Matlab code block provided in the article.

    X = [ 1 2 3 ];
    Y = [ 1 2 3;
          4 5 6;
          7 8 9 ];
    Z = Y * X';
    W = [ Z Z ];

I'm not sure I understand your point. You're saying that channels are like apt/sources.list or yum.repos.d, and flakes are like the Apple App Store? Or the other way around?

One thing that probably didn't help my understanding of channels was that I run Nix on non-NixOS systems (primarily MacOS and Fedora). If I'd stuck to NixOS, then thinking of a channel in the same terms as apt/sources.list or yum.repos.d would have been an easier mental model.

> The flakes were the main UX/DX improvement for me. Before them I honestly could not do anything.

Agreed. I think flakes are far more intuitive than channels. In a flake everything is declared in the repo it's used in. I still don't understand channels.

For someone who's used to thinking in channels, I suppose flakes would be jarring. For someone (like me) who came from the world of Project.toml and package.json, flakes make a lot of sense.

I've started experimenting with Typst for a few documents, and here's my stack:

- Zed editor with Typst plugin

- Tinymist LSP settings turned on to render on save in Zed, see https://code.millironx.com/millironx/nix-dotfiles/src/commit...

- Okular open to the output document. Okular refreshes the document when changed on disk.

It's not as polished as say, LaTeX Workshop in VSCode, but it gets the job done.

Bureau of Labor Statistics says that the mean annual wage of "Automotive Technicians and Repairers (SOC code 49-3020)" is $55,780 as of May 2025, so yeah, something doesn't add up.

I remember hearing about that rumen swap experiment. One of my professors at the time said that more substantial changes could be affected by doing repeat rumen content swaps.

Hugo includes native KaTeX: https://gohugo.io/functions/transform/tomath/

The docs recommend setting up KaTeX CSS (which requires either a CDN link or Node), but by changing output to 'mathml,' you can have the browser render equations with zero dependencies.

> Is this normal for the machine?

No, it's not "normal," but it is fairly common. When I worked in NGS, nearly 1/4 of flow cells were duds. ONT used to have a policy where you could return the cell and get a new one if it failed its self-test.

I think all elected officials plug their ears when they hear this.

I know the governors of Wyoming gave land away to NCAR and Microsoft talking about how they were going to "diversify Wyoming's economy," and then every employee of those data centers was a contractor out of Colorado. The current governor has kept up the hype, now claiming that AI will be good for the local energy companies despite the fact that the proposed centers are going to be connected to out-of-state energy pipelines.

When you look at top campaign contributions and see Google and Microsoft at the top, you understand why the gubernatorial class keeps their ears plugged.

> scientists have uncovered a mechanism that could help explain this connection, finding that bacteria can travel through swallowed saliva into the pancreas

Except, this study didn't do that. It did shotgun sequencing and found a correlation between certain microbial species (some were fungal, not bacterial) and cancer risk. It *did not* demonstrate anything about mechanism.

Based on the way it's phrased, maybe this article is saying that previous studies have found a mechanism, and this study found the microbial culprits. Unfortunately, I don't have access to the full study to see if that's the case in its introduction or discussion. Even so, that's an incredibly misleading opening to the article.

Was this in any way inspired by Fireship's recurring gag, "Tinder for Horses?"

Except Apple code signing on MacOS is basically what Google is trying to copy over to Android. I can run arbitrary programs on MacOS, but I have to go and remove the com.apple.quarantine attribute from any application that doesn't have Apple's explicit permission to exist, i.e. most FOSS apps. I suspect that option will go away eventually.

I don't have any great ideas on how to do it right - I'm in the "don't know what I don't know" phase of electronics. Part of the reason I quit playing with electronics was that I already knew Visual Basic pretty well, and so I could program engaging things quickly while the electronics took a lot of fiddling with no understanding to sometimes get a result.

This has been a handicap in my career where I would by default reach for a Raspberry Pi or a National Instruments RIO in places where an IC or a Arduino would suffice. Stuff like data acquisition units (DAQ) and control systems are prime examples. Maybe by giving those applications first? I can think of several projects my 12-year-old self would have found a use for a DAQ or control system.

Cool idea. I grew up with the most basic Snap Circuits kit[1] and a solderless breadboard kit from RadioShack[2] - other toys you might look to for inspiration.

I think I was about 12 when I got these, and I remember that the learning curve between them was pretty steep - I was building all sorts of custom circuits with Snap, but had maxed out the capabilities of such basic parts pretty quickly. I never did figure out how to make anything other than the step-by-step projects with the breadboard kit. Although the ICs on Snap are kind of laughable, I could at least figure out what they did (also I seem to remember every one of them just played a sound), but the Radio shack kit never really explained why anything worked the way it did or what the applications were outside of its recipes, so eventually it just went on the shelf and was forgotten. That would be one trap to avoid from my experience.

[1]: https://shop.elenco.com/consumers/snap-circuits-jr-100-exper...

[2]: https://www.amazon.com/RadioShack-28-280-Electronics-Learnin...

You can use Antidote to selectively load the parts of OhMyZsh you need - https://github.com/getantidote/use-omz

My own usage: https://code.millironx.com/millironx/nix-dotfiles/src/commit...

> tests on dead pig organs do not test the robots’ capacity to react to a patient moving and breathing, blood running in the field of operation, an inadvertent injury, smoke from cauterisation or fluid on the camera lens

Really cool tech being demoed here but calling it "surgery" is a bit of a stretch considering how much of surgery time involves dealing with hemostasis, and "the dead don't bleed."

The article does go into this.

> But what sets Spielberg apart from Hammond, what sets Jurassic Park apart from it’s imitators—and why the film industry now paradoxically needs Spielberg after he helped to weaken it—is that Spielberg had the discipline.

I would say that's the crux of the argument: the reason there are no good dinosaur films is that there are no people with Spielberg's discipline to make them.

One can try to blame studio interference, investors, streaming, the latest political trend, etc, or even Spielberg himself for why there are no disciplined producers left in Hollywood, but that's not what the article was about.

Usability, for one. I know how to `apptainer run ...`. I don't know how to do what you're describing, and I'd say I'm more Linux savvy than most HPC users.

Server - but this is kind of a wrong question. Apptainer is for running cli applications in immutable, rootless containers. The closest tool I can think of is Fedora Toolbx [1]. Apptainer is primarily used to distributing and reusing scientific computing tools b/c it doesn't allow root, doesn't allow for changes to the rootfs in each container, automatically mounts the working directory and works well with GPUs (that last point I can't personally attest to).

[1]: https://docs.fedoraproject.org/en-US/fedora-silverblue/toolb...