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

A developer created Lucid, a web tool that visualizes what language models are "thinking" before they generate responses by reading internal representations layer-by-layer using the Jacobian lens technique. The tool allows users to input prompts and observe which concepts the model activates at different processing stages, revealing both what the model considers and what it ultimately outputs. The public version works with small open-source models and requires no account or installation to use.