wanna know the cs fundamentals you need to actually understand LLMs?
- just ~1.5 years or less of focused work
- no fluff, no degree, no gatekeeping
- here's the map:
step 0: python
- learn by building: scrapers, games, basic neural nets
- vibe with numpy, pytorch, matplotlib
- everything downstream speaks fluent python
step 1: data structures & algorithms
- arrays, linked lists, stacks, queues
- trees, graphs, hash maps
- sorting/searching — not for interviews, for intuition
step 2: discrete math
- logic, sets, functions
- combinations & permutations
- graphs & probability — sneakily essential for LLM reasoning
step 3: computer architecture
- bits, bytes, binary, memory hierarchy
- what is a floating point number
- how CPUs/GPUs chew through matrix math
- understand hardware so you can bend it
step 4: operating systems & networking
- threads vs processes, memory management
- sockets, HTTP, DNS, latency
- how LLMs talk to each other at scale
after that:
- LLM internals finally make sense
- tokenization → embeddings → attention → logits → sampling
- no longer magic, just math
this path takes ~1.5 years
- no CS degree
- no expensive bootcamp
- just curiosity, consistency, and a strong tab game
the elite don't want you to know this
but now you do;
so do the work,
put in the time,
and before you know it:
you'll be the one building the models