Show HN: How LLMs Work – Interactive visual guide based on Karpathy's lecture | Hacker News
TL;DR AI
2 min readKey summary
A Hacker News post shared an interactive visual guide based on Karpathy’s lecture on how LLMs work.
The demo explains variable-length token inputs, including padding and masking for sequences of different lengths.
It also shows how embedding layers turn tokens into vectors and how context disambiguates meaning.
The guide ends by illustrating the projection back from hidden representations to token space.
Same topic
