update: i am teaching transformers to a bright high schooler
from no calculus / linear algebra to transformers, all notes taken on my own ai-live-edtech software (infini) !!!
update:
>lost my website code
>made a simple webserver in C that literally serves a text file
>platforms can't host it
>setup custom vps on hetzner
>setup traefik
>setup portainer
>setup custom registry
>custom github action
>infinite dockers
>one click commit txt to
update: calculus v2
thought of a better way to explain calculus 1. derivative as generalized slope is more intuitive imo
it's the funniest thing that a 16y/o girl is doing college intro lv math for jiggles hahah
mini-make-monday
I love doing market research just to find a YC company that does exactly what I want, so I worked all day to share the joy with the world!
startup idea -> existing YC companies !!!
Find YC companies that do exactly what you had in mind:
update: calculus v2
thought of a better way to explain calculus 1. derivative as generalized slope is more intuitive imo
it's the funniest thing that a 16y/o girl is doing college intro lv math for jiggles hahah
update: speedrunning calculus
how do you best explain the intuition behind functions? limits? derivatives? it's easy to forget how it felt not knowing what dy/dx meant
ml truly from first principles
update: speedrunning calculus
how do you best explain the intuition behind functions? limits? derivatives? it's easy to forget how it felt not knowing what dy/dx meant
ml truly from first principles
update: i am teaching transformers to a bright high schooler
from no calculus / linear algebra to transformers, all notes taken on my own ai-live-edtech software (infini) !!!
update: i am teaching transformers to a bright high schooler
from no calculus / linear algebra to transformers, all notes taken on my own ai-live-edtech software (infini) !!!
📽️ New 4 hour (lol) video lecture on YouTube:
"Let’s reproduce GPT-2 (124M)"
The video ended up so long because it is... comprehensive: we start with empty file and end up with a GPT-2 (124M) model:
- first we build the GPT-2 network
- then we optimize
Karpathy Books Update:
- Nexus 👍 Neuralink turned drug, fun, feels like reading an action movie
- Permutation City 👎 weird in a not very interesting way, the plot is all over the place, feels like cartoon scifi
Running dynamic WASM is god awful, there must be a better way to load / instantiate / execute modules at runtime without all the memory mangement from JS, no?
like allocate a bunch of memory? never use the heap? single module embed? it's all just dumb?
# explaining llm.c in layman terms
Training Large Language Models (LLMs), like ChatGPT, involves a large amount of code and complexity.
For example, a typical LLM training project might use the PyTorch deep learning library. PyTorch is quite complex because it implements a very
@mrsiipa
for a short intro, feynman lectures (vol 3?). griffith's intro to quantum mechanics is goated, very easy to follow. wouldn't suggest dirac's / landau, they're pretty math heavy. but i swear a good textbook is far better than professors
@notmoeezm
we keep pushing the abstractions - variables, complex numbers, matrices were once considered highly abstract and now they're standard material. pretty sick to think that all algebra was abstract at some point
Meet the new iPad Pro: the thinnest product we’ve ever created, the most advanced display we’ve ever produced, with the incredible power of the M4 chip. Just imagine all the things it’ll be used to create.