Understanding LLMs From Scratch: Tokenizers

You wake up, Rip Van Winkle-like, from a deep sleep. It’s 2026 and LLM’s are taking over the world. Time to learn something about how these work, you think (and write up your notes). Best to start at the very beginning of the LLM processing pipeline - what is a tokenizer and how does it work? What are Tokenizers? At its most basic, a tokenizer is a way to take a string of text and turn it into a list of numbers that a computer can understand. ...

August 20, 2026 · 4 min · Sidney Fletcher

Rust - In-Place Init for pin-heads

The Rust project has been working on in-place construction of values. I’m not much of a PL person or language designer but as someone who’s been following the debate, I thought it’d be fun to write up a small variant on what’s being discussed! Basic Proposal The basic idea I had is a hybrid of Yosh Wuyts’s “placing functions” proposal and Olivier Faure’s pinned places ideas that I’d call “return position pinned r-values” (RPPR)1 ...

6 min · Sidney Fletcher