Lucid - Visualize AI internal concepts before answers

Show HN: I built a web tool to see and edit what an AI thinks before it answers

Lucid is a groundbreaking web tool that lets you peek inside a language model's mind before it generates a response. By utilizing the Jacobian lens, this instrument reveals the hidden concepts and internal representations an AI holds layer by layer, effectively mapping the space between input and output. Users can type any prompt to watch which ideas enter the model's global workspace, identifying thoughts the model considers but never vocalizes. Running on open models like Qwen and Pythia, Lucid requires no account or installation, offering an immediate, interactive way to explore the cognitive mechanics of AI and understand the 'J-space' proposed by Anthropic.

"Lucid watches a language model think. It reads the concepts a model holds before it speaks, layer by layer, using the Jacobian lens."