← back

ai playground 🧪

small interactive demos of the primitives behind every LLM product. everything runs in your browser — view source to see how.

01 · tokenizer visualizer

before a model reads your prompt, it gets chopped into tokens. type below and watch how.

loading tokenizer…

02 · embedding similarity

embeddings turn text into vectors so "meaning" becomes measurable — the core trick behind RAG and semantic search. compare any two sentences.

this demo downloads a real embedding model (all-MiniLM-L6-v2, ~25 MB) and runs it entirely in your browser — no server, no API.

⌘K