Windows Server 2025 is somewhat better on reads but only at low parallelism.

There’s no difference on writes.

Even someone who has never seen a Windows PC in their life could guess what this script does.

Linux and Unix shell commands use completely arbitrary single letter parameters that must be looked up or memorised. That’s not a virtue.

[0]: https://learn.microsoft.com/en-us/powershell/module/microsof...

Contrary to how it sounds I actually like PowerShell as a scripting language in itself. A lot of its ideas are pretty clever.

I treat my dormant familiarity with it as a resume hedge. Ideally things in my life continue to go well, and I can keep climbing the ranks of doing more and more impressive things powered by the Unix systems I've been daily driving since I was 14. If, however, things in my life ever go truly sideways, I could probably dial it way back and eke out an existence at some low pay, low stress, fully remote Windows admin job at some dinosaur of a company somewhere. There I could use PS and its deep, deep integration with all things Windows to automate 90-99% of it away, so that I could spend my time e.g. tending to my young children instead. (Even if Copy-Item is 27% slower than drag and drop. My time is still more expensive than the machine's.)

I truly never hope that has to happen, of course.

Lithium, rare earth metals, and a bunch of others are only "scarce" because right now China is the only country willing to put up with the pollution levels that the cheap, dirty version of their extraction produces.

Everything can be produced cleanly, safely, etc... but that comes at a price.

It's like when employers complain that "nobody wants to work". That needs to be translated to "nobody wants to work for the low wages I'm willing to pay".

Mathematica is way, way under appreciated in industry, and even in the sciences.

That's essentially what's going on with AI models, they're struggling because they only get "one step" to solve the problem instead of being able to trace through the maze slowly.

An interesting experiment would be to ask the AI to incrementally solve the maze. Ask it to draw a line starting at the entrance a little ways into the maze, then a little bit further, etc... until it gets to the end.

https://arxiv.org/abs/2407.01392

of course it doesn't redraw the image on every step, so not exactly what you're suggesting (interesting idea btw) but i think it's relevant.

Now you can be mad about two things nobody else notices.

That just gave me an idea! I wonder how useful (and for what) a model would be if it was trained using a two-phase approach:

1) Put the training data through an embedding model to create a giant vector index of the entire Internet.

2) Train a transformer LLM but instead only utilising its weights, it can also do lookups against the index.

Its like a MoE where one (or more) of the experts is a fuzzy google search.

The best thing is that adding up-to-date knowledge won’t require retraining the entire model!

This has been thoroughly studied and is well known to geologists, especially in Northern Europe where changes can occur over human time scales.

I remember watching a documentary about a landslide disaster that occurred because clay saturated by salt water had lifted above sea level due to GIA and the fresh rainwater had washed the salt out. Clay without salt ions is more liquid and eventually moved catastrophically.

Stay Healthy!

.. and pretty much why I worded the first line of my comment above as I did.

Sea level changes, glacial retreats, etc will all change crustal pressures and impact the frequency, placement, and nauture of volcanic activity in times to come.

Not a lot of that in Ethiopa for now though (to the best of my current "haven't specifically looked at the map for this" gut feeling).