Live demo
A streaming chat, faked end to end
On DDMind and CCM I built WebSocket chats with streaming responses, so users could ask a question about a company, report or contract and read the answer as it arrived. This demo recreates the front-end mechanics — token-by-token rendering, auto-scroll, Stop, error and retry — with a fake server and canned answers about my own work. No AI API is called and nothing is sent anywhere.
Ask about my work
Pick a question below. Answers stream in token by token, like the WebSocket chats I built on DDMind and CCM.
How this maps to the real thing
In production the stream came over a WebSocket from a backend that routed between OpenAI and Anthropic. The UI concerns are identical to what you see here: buffer tokens into a message, keep the view scrolled unless the reader scrolls up, let people cancel a long answer, and recover cleanly when the connection drops. Read the DDMind case study for the wider architecture.