Not tech related, but the advice from Jim Donovan, a very smart and successful Goldman Sachs exec with a chemical engineering BS and MBA from MIT and a JD from Harvard contains a lot of alpha and similarities to Lud's advice here
implemented a bare bones x86 32-bit kernel in Zig
zig's build system is so powerful, it handles all the linking and even runs qemu
going to continue development on the kernel in parallel with my reading of OSTEP
@ludwigABAP
@__tinygrad__
shifted my obsession from ML to compilers and systems programming so fast
realizing tensor libs are basically just just DSLs with a python front end blew my mind and opened up a whole world i had to dive into
@0xluffyb
your entire lineage that survived catastrophe after catastrophe and pushed and sacrificed for you to be in the position you are this very moment
do people ever get to a point where they can look at code they wrote a year ago and still think its good?
is there a point where u can max out ur programming skills?
i think the solution is to look at good codebases and read stuff from matklad and all the tigerbeetle ppl
for example matklad’s “push ifs up and fors down”
rustcels have made it past denial and are now in the anger stage of the grief process
i urge my fellow zigglers to remain patient when interacting with these ppl, it’s difficult to cope with a better language popping up
@justalexoki
I can’t tell you why it’s set up that way but as an American it’s bothersome to see people come to my country and not participate in our customs
@MomentumOp89
You feel that way until it pays off. The long term delayed gratification is real.
It paid off for me a little bit last month, and the happiness I felt after grinding to earn it was amazing
"Andrew Kelley - Practical Data-Oriented Programming"
Founder of Zig talking about an optimization approach for efficient usage of CPU and memory caches
why do people so willingly confess to not understanding pointers?
not something that should ever be accepted unless you’ve been programming for <1 yr maybe
its the programming equivalent of people saying they’re not a “math person”
is there anything more pathetically emasculating than compiling someone else’s and running it on YOUR MACHINE?
take some computational responsibility. your machine is a temple, and ur just going to let somebody else manipulate it?
ur going to let somebody else scramble ur
when i was younger you would get made fun of online for asking stupid questions, get sent let me google that for you links, and lose cred
now all anybody knows is ask chatgpt and immediate spoonfeeding from peers
i have to tell my roommate (whos learning to code) to just google things every single day. like you can just talk to the fucking pagerank graph of the entirety of human knowledge and you choose to flounder and say you DONT KNOW
@pixqc
when we write assembly we don’t have to worry about managing cache the same way when we write c we don’t have to worry about managing registers
assembly uses the cpu’s instruction set to interact with the cpu, which itself has to manage its cache
i just meant that there’s
i get so excited when i learn that a moot is younger than me
can’t imagine how cracked someone that spends time in this twitter ring as a teenager will be at my age
is there anything more pathetically emasculating than compiling someone else’s and running it on YOUR MACHINE?
take some computational responsibility. your machine is a temple, and ur just going to let somebody else manipulate it?
ur going to let somebody else scramble ur
the reason “just build stuff” is so important is that positive momentum compounds and your confidence to tackle new problems skyrockets
as you gain more general knowledge and confidence, you find opportunities to innovate and build in places you usually wouldn’t
got some simple tensor operations calculating at comptime using zig's built-in vectorization/simd
still super basic but i'm gonna see if i can keep building up to a comptime neural net
luddite "coder" too purist to adopt new tech, stubbornly sticking to manual coding, getting left in the dust, obsessing over syntax
vs
modern programmer embracing LLMs, getting things done, making money, solving problems, building solutions
If you're against LLM-assisted programming, you're ngmi. we're programmers and builder who build cool shit. we're not "coders".
coding is a primitive, mind-numbing task. you should be instulted if somebody calls you a "coder". coding is just translating human ideas to something
this is much, much cleaner
we now allocate a single large block of memory during model initialization, which is then partitioned into sections for parameters, gradients, optimizer states, activations, and their gradients
paramatertensors and activationtensors are structs
If you're against LLM-assisted programming, you're ngmi. we're programmers and builder who build cool shit. we're not "coders".
coding is a primitive, mind-numbing task. you should be instulted if somebody calls you a "coder". coding is just translating human ideas to something
probably going to regret my lack of foresight and misplaced priorities during my teens for a long time, at least until i feel like im at least somewhat close to reaching my potential
i unknowingly cut off so many opportunities before i fully understood their value
forward pass is done
now we need to implement the update and backward pass functions and the model will be finished
still a long way to go until we're training on datasets after that though
zig enables you to format a struct for efficient memory layouts or specific implementation requirements and then use namespaced functions to provide clean interfaces to interact with
ex: entry struct for an interrupt descriptor table with a specific required memory layout
@shauseth
You don’t get ahead by just being gifted. You get ahead by applying your talents tirelessly, working to maximize your natural skills, and taking the initiative to use that talent to add value.
something so magical about functional programming i can’t put a finger on it
every time i write something in an FP i get the same feeling as when i first read SICP and was in awe of the beauty of computation
learning very quickly with zig that taking time to properly plan and design the system will save you so much time down the line during implementation
tiger style was always right
implemented a bare bones x86 32-bit kernel in Zig
zig's build system is so powerful, it handles all the linking and even runs qemu
going to continue development on the kernel in parallel with my reading of OSTEP
“Don’t be a boilerplate programmer. Instead, build tools for users and other programmers. Take historical note of textile and steel industries: do you want to build machines and tools, or do you want to operate those machines?”
- Ras Bodik
@ronawang
imo
programming will further bifurcate into extremes of high and low-level. developers will be extraordinarily productive, but will understand less about how their computers work. systems engineering will quietly determine the bounds of what everyone else builds
Never ask a question you could find the answer to yourself. Answering your own questions is the only way to learn and asking inane questions is disrespectful to the questioned’s time.
Learn how to ask good questions
when i was younger you would get made fun of online for asking stupid questions, get sent let me google that for you links, and lose cred
now all anybody knows is ask chatgpt and immediate spoonfeeding from peers
@Jorge_DBG
its a great book its what im currently reading. i think its a must read for anybody that wants to work on or build systems that support a lot of users/data
@ludwigABAP
I thought I was dunking on a javascript dev but it’s actually someone who works on rust full time and has a three part blog series going deep on pointers
there’s a unique pleasure in writing a piece of software where perfection isn’t a goal, but a necessity.
where every element demands precision. each byte must find its exact address. each component must harmonize into a flawless symphony.
if you are not humble and take accountability for yourself you are not free
blaming external factors is a fatalistic. if you believe somebody sees more success because they were given more opportunities or because they have more innate god-given talent that you don’t have, you
New Jersey, being the home of Bell Labs, is where the transistor, Unix, C, and information theory was invented.
10 Nobel Prizes and 6 Turing Awards have been awarded for work done at Bell Labs in New Jersey