One use case I can see AI glasses could solve that no other devices can is contextual reminders - when I eat, leave home, etc. This could be a game changer for elderly with memory issues too. Of course personally I would only consider using that with local processing.

Would be great if CSS allowed to refer to the computed element size in properties that cannot affect said size, i.e. transforms, animations and transitions. Maybe I should use this an excuse to dive into @ladybirdbrowser and build a prototype 😅.

I think React hooks may have been a net negative. They do provide a very powerful API, but: - there are way too many gotchas - it is too easy to write spaghetti code - if I want to less rerenders, my code becomes a wall of useMemo / useCallback

Tried bun.sh: > bun test segmentation fault I guess it needs to bake some more… 🙃

I often consider dropping @WebStormIDE in favor of something more lightweight, but it’s static analysis is still way superior to what anything else JS/TS related offers. Saved my butt more than once. I only wish I could run its checks on CI, not just in the IDE.

Programming tip: If you make an API to do something, and you can imagine people needing to call it more than once, it should probably have a bulk version. Or even only have a bulk version. This is especially important for IO workloads including HTTP, RPC, DB, etc.

It has been a while, but I just published new compiler programming video. Continuing my from-scratch language and a compiler written in C. This time working on a built-in debugger: youtu.be/O2eRJSdFoDE

I just wanted to write a small exploration project to better understand Lottie animation file format. 3 levels deep in yak shaving and I am writing a compiler for matching expressions in AST-like data structures 😅

I now have very mixed feelings about the optional chaining (foo?.bar) in JavaScript. It is useful occasionally, but at the same time it makes it way too easy to have loosely typed and validated data floating around throughout the whole codebase 🤔

Every few years I try to make a code editor (syntax highlighted) in a browser that works without a custom text input and undo / redo system like CodeMirror, or VSCode do. So far it always turned into a hackfest and ultimately failed. Going for round 6 or 7 now... Wish me luck🍀

My JavaScript parser can now also print the source code (with poor formatting) and parse it again with the identical AST. 🎉 This is surprisingly harder than it sounds because of all the dark corners of the JS grammar. Now I'm going to indulge myself in some more optimization.

JavaScript regex literal syntax is a cautionary tale on how a small feature / decision can completely screw you over in a long run. Even C/C++ typedefs aren't as bad for parser writers.

I wonder if I know someone who could say why both Windows and Mac OS updates take 10s of minutes to complete. In that time you could overwrite full capacity of the SSD dozens of times. This is not a troll tweet - I actually want to know.

Been spending a few evenings on writing a JS parser and all I can say is that the language has quite a few dark corners. And haven't even touched the whole regex literal mess yet. Progress is pretty good though (>50% of test262)

TIL that in JavaScript you can put property access part of the method call in parens and will still get the correct "this" 🤯 ("foo".startsWith)("foo") is the same as: "foo".startsWith("foo")

Got contacted by a recruiter from a MAANG company. The logo in the email weighs almost 2mb and has 29000x12500 size scaled down to ~200 logical pixels. You know, just in case you have a 128k screen from the future.