Case Study — Educational Tool
Perceptron LLM
An interactive playground for watching how a language model actually picks its next word — one probability at a time.
What I Did
Type a prompt and Perceptron LLM generates it one token at a time, surfacing the candidate next-words and their probabilities at each step — with temperature, top-p, and max tokens exposed as live controls, so the usually-invisible sampling process behind an LLM's output becomes something you can see and tune yourself.
