Ronald Chen (@pyrolistical) header
Ronald Chen (@pyrolistical) profile picture

Code Powerleveler

Joined July 2015
0100 eat 0800 sleep 1530 wake
1
wat > new Uint8Array(new ArrayBuffer(100), 0, 10).length 10 > new Uint8Array(new Uint8Array(100), 0, 10).length 100
1
83
Colorblind users should config their git with [color "diff"] old = red reverse new = green reverse
64
I wish @ziglang splat was a little smarter
1
200
Is chat gpt just bad improv?
1
2
62
VR people: One day we will be able to enter the metaverse Apple: We are bringing the metaverse to your living room
1
1
1
66
As hardware perf per dollar is stalling, it makes business sense to spend money optimizing existing code to run faster. Optimization can yield exponential improvement where as buying more hardware is a linear cost. Most code was never written with performance in mind
1
44
We can squeeze so much more out of our existing hardware. Remove layers: software stacks, operating systems, containers, virtual machines, frameworks Run code faster: SIMD, parallelize, fixed function accelerators Use memory better: use less, cache better, maximize NUMA
34
How to refactor code quickly with tests: 1. Tests are required to exist 2. Tests are required to be to good Neither of which are generally true, so we end up not refactoring at all
36
We want a distributed system to be globally sound. But event driven services are only locally sound. It’s hard to think about the global consequences when all other event/services are asynchronous/changing/down
39
AI will save the world by destroying its biggest threat, the humans
39
$ cat ~/.zshrc PS1='%2d$ '
25
Don’t tell me no, tell me the trade off
33
Is burnout the creeping realization what you are doing is worthless?
41
We already have all the tools we need to make anything. It would be a shame if we just wasted all our time and produced nothing at all
1
43
Why use an UI when I can spend double the time writing code to do the same thing?
28
If we want AI image generation to draw hands well, we need to give it hands
38
Oh god. Learning rate is a numerical analysis problem
28
It is silly to compare the dangers of artificial intelligence to nuclear research. Anybody with a computer can build artificial intelligence. On the other hand entire nations are required to refine Uranium
42