what surprised me is how different the rendering architecture is for each framework.

Win32 -> message loops & GDI

Winforms -> managed C# via P/Invoke

WPF -> throwse all away and uses DirectX

UWP -> Appcontainer sandboxing

WinUI -> decouples from OS entirely

This visual breakdown helped me to see it clearly - https://vectree.io/c/evolution-of-windows-gui-frameworks-fro...

As far as I remember, WinForms is just a thin wrapper around Win32 with a message loop, i.e. not all that different.

So.. eat more fiber and your gut bacteria will produce butyrate. And that helps muscles.

The article shows WASM being 1.2-3.7x slower, and your experience confirms it.

Do you have any idea which operations regress the most?

Based on looking at V8's JITed code, there seemed to be a lot of overhead with stack overflow checking, actually. The function prologues and epilogues were just as bloated in the tail-call case. I'll upload some screenshots if I can find them.

Chain of thought happens in the <think> tags, not the visible output.

Caveman only strips filler from what you see... the reasoning depth stays the same.

I found this visualisation pretty interesting - https://vectree.io/c/chain-of-thought-reasoning-how-llms-thi...

Great find! It appears uppercase tokens are completely unknonw to the tokenizer.

But the character still comes through in response :)

The wiki diagram helped me too, thanks!

One thing I did not understand before: why SNAT must happen at POSTROUTING?

Because the exit interface is only known after the routing decision... before that kernel does not know which source IP to write

this visual schematic made it click for me - https://vectree.io/c/linux-netfilter-packet-flow-tables-chai...

Good point, full self driving probably felt like marketing win back then... but now it is a liability. Rebranding now would just highlight all missed promises

Looked at their source code out of curiosity. They use Elasticsearch as a geo backend, tile server that renders PNG images server side for each map tile request.

- low zoom -> server aggregated grid

- high zoom -> switcher to point tiles

browser just displays images, there is no work to do on client side..

I generated a visual schematic how these kind of systems work: https://vectree.io/c/server-side-geo-tile-rendering-elastics...

yes, this part I find to be most interesting.. how many losses before the picture changes for the public?

maybe it keeps getting rediscovered, becaues a good tooling around it is still missing?