I noticed I find it impossible to turn off when coding (or writing in general). I wonder if, when working on a very visual problem, there's a cognitive context-switching performance overhead from constantly having to switch from visual to inner-monologue thinking and back

Show this thread

zack

@zack_overflow

Oct 5

Meditating has given me a greater control over my inner monologue, it helps me tame the signal/noise ratio of my consciousness It also has improved my visual cognition, inner monologue gets in the way of visualization if you can't control it well

Do you have an inner monologue? If so can you turn it off when you code? Or do you think in abstract thoughts or visions/images/colors and what is that like when coding?

Here is Ezno, the compiler thingy I have been working on:

kaleidawave.github.io

Introducing Ezno

Introducing Ezno and the current state of the project.

203

931

Show this thread

zack Retweeted

Bug fixes & performance improvements

@humanly_typed

Sep 21

Replying to

@stylewarning

Dynamic typing gives you that dopamine hit quicker; few hundred lines of code later, it drowns you in cortisol.

Delighted by

new "Val" language val-lang.dev Feels close to the sweet spot of comprehensible ownership semantics I often daydream about having kept Rust at ("no 1st-class borrows or lifetimes, only 2nd-class parameter modes, plus yield")

val-lang.dev

Val

The Val Programming Language

262

dis funny: PoAC has been successfully implemented, but for only 9 packets of data, with a packet loss ratio of 55% (due to operator error), and a response time ranging from 3,000 seconds (50 min) to over 6,000 seconds (100 min). Thus, this technology suffers from high latency.

wtf, just learned about IP over Avian Carriers... using homing pigeons to deliver internet protocol packets

en.wikipedia.org

IP over Avian Carriers - Wikipedia

This is a cool paper about achieving cache friendliness with abstractions that feel high-level enough that they're not error prone / cause less logical errors. It also has a title ("You Can Have It All") that sounds like a 80's pop song doc.ic.ac.uk/~scd/ShapesOnw

zack

@zack_overflow

Aug 30

Super interesting because one would expect the opposite!

Mercury's CTO

attributes it to: • more demand, less supply • good developers tend to be interested/know Haskell ("Interest in Haskell acts as a decent proxy for baseline developer quality.")

"[...] our VP of People has previously said that backend Haskell engineer is the “easiest role to hire for” in the entire company, and in my opinion hiring is Haskell’s “killer app”."

rlly Interesting that

@mercury

has an easier time hiring devs for Haskell (a very esoteric lang) than TS devs "Too many companies recruit for common languages. Our recruiters have to send 10 times as much outreach to get a response for TypeScript jobs."

serokell.io

Haskell in Production: Mercury

In this edition of our Haskell in Production series, we interview Max Tagher, the co-founder and CTO of Mercury. Read the article to learn where Mercury uses Haskell, why they chose it, and what they...

I present to you submillisecond, a Rust web framework for WebAssembly backend apps. Powered by lunatic!

github.com

GitHub - lunatic-solutions/submillisecond: A lunatic web framework

A lunatic web framework. Contribute to lunatic-solutions/submillisecond development by creating an account on GitHub.

zack

@zack_overflow

Aug 24

Found out my TV has a "Dialogue enhancer" setting, no clue if it's audio compression though

I know most modern TVs have multiple settings for adjusting picture quality / colors, are there similar settings for adjusting audio compression to hear dialogue better?

In audio engineering, compressors (not to be confused with data compression) reduce a signals dynamic range, reducing the difference in volume between the loudest and quietest sounds Whenever I watch movies/tv on VLC I turn up the compression to hear dialogue better

Quote Tweet

Zack Bornstein

@ZackBornstein

Aug 22

Am I going deaf or has anyone else had to watch every TV show with subtitles since like 2019

Show this thread

Design patterns every programmer should learn in the 21st century: 1. Associativity 2. Currying 3. Sum types / Product types 4. Fold / Map 5. Passing functions as arguments 6. Immutable data structures

147

866

zack

@zack_overflow

Aug 13

Anyone know if this exists already?

I wanna make a new Vim mode called "syntax mode", when enabled you can manipulate code on the syntax/ast level (i.e. structured editing) And a popup appears above the token you're editing (like the intellisense popup) and shows the token's place in the syntax tree

Really cool obscure JavaScriptCore feature: DOMJIT Embedders can manually JIT native functions on a per-function basis This is the code Safari uses for JIT'ing document.firstChild github.com/oven-sh/WebKit

wtf almost 2 years ago I made a site and literally forgot about it until just now just logged into google analytics and i see it gets 500k impressions, 25k clicks, and 37k users a month!

Synapse ⚡️ It’s a neurone to neurone connection, or in other words, small gap at the end of a neurone that allows impulses to pass from one neurone (presynaptic) to the other (postsynaptic)🧐 An amazing Video to visualise how a synapse works😍 - Credit of Video:Helix animation

0:21

1.7M views

506

1,905

Show this thread

zack Retweeted

Rev. Dr. Everett Howe

@howe

Aug 7

“In high school we learned that 2H₂ + O₂ → 2H₂O. It’s ridiculous to think that that simple reaction is sentient. So how could a human brain, which is simple a large system of such reactions, produce thought?”

Quote Tweet

Noah Giansiracusa

@ProfNoahGian

Aug 7

Wrote op-ed w/ Nobel Prize winning economist @paulmromer in which we explore the power of mathematical functions (and mathematical thinking) to provide a clear debunking of AI sentience and to reveal what's been largely missing from the debate.

1/14 barrons.com/articles/ai-ar

Show this thread

thanks to

for sending me this article which is is topical and appealing to fans of breaking bad and better call saul

604

Buuuut that proposal is inactive now, in favor of the component model proposal, which currently doesn't have a solution for this written in the design but lists "adapter functions" as a future feature. So it could be a very long time before its ready.

The end goal is that browser could also implement these, so instead of converting from WASM -> JS -> DOM you directly go from WASM -> DOM

But this could be changed in the future WASM's Interface Types proposal addresses this exact problem. It lets you define an "IR" so you aren't stuck with WASM's fixed representation of data types

I think the only compelling reason to use a WASM web framework right now is if you want to write code in an another language: Rust/Zig/C++/etc. Which is valid, I just don't expect something like Yew to be much faster.

So I don't think perf is a super compelling reason to use a WASM frontend framework. And that's also why rewriting React/Svelte/SolidJS/etc. in WASM won't really amount to significant performance gains, if any.

Keep in mind wasm-bindgen is the lowest-level possible Rust WASM framework, it just calls vanilla DOM functions from WASM. It's the vanilla JS equivalent in WASM

That's why in the JS framework benchmark (github.com/krausest/js-fr), vanilla JS always outperforms wasm-bindgen Except in startuptime cause WASM bytecode is more compact than JS source code and JS also needs to be parsed and compiled into V8/JSC's bytecode

github.com

GitHub - krausest/js-framework-benchmark: A comparison of the perfomance of a few popular javascr...

A comparison of the perfomance of a few popular javascript frameworks - GitHub - krausest/js-framework-benchmark: A comparison of the perfomance of a few popular javascript frameworks

It's particularly bad with strings, not only do you have to copy them, but convert them to several different representations (ex/ JS UTF-16 -> WASM array of numbers -> Rust UTF-8) Web frameworks read/write many strings (text content, classes, style attributes, etc.) from the DOM

But in a WASM frontend framework, you havee to call the DOM through imported JS functions (there's no way to access the DOM directly from WASM), so there's no escaping this performance cost

There is always an overhead from copying/serializing/deserializing data from WASM -> JS and vice-versa and it can be bad for perf. For WASM to outperform JS you kinda have to minimize how frequently data crosses the JS/WASM boundary, to reduce this redundant copying/computation

Rambling about why WASM web frontend frameworks can't be faster than vanilla JS frontend frameworks (at least right now):

275

Really surprised Rust/Zig don't do this, generics/comptime make this feasible... right? Is there something I'm missing?

Why don't Rust/Zig/etc make their array list type parametric over the amount of bytes used to store the capacity & length? Most langs use u64s for the length & capacity, which can be wasted space / bad cache locality if the item count is small

The above example goes beyond type-safety, it's a performance optimization. You only need to check the array is sorted once and then you can use it in other functions that expect a SortedArray. Another example, composing refinement types for even stronger types:

Show this thread