Lucid
Lucid watches a language model think.
What you get: Shows which concepts a small language model holds internally, layer by layer, before it produces its answer.
Score out of 100. Mostly what the app is worth to a visitor,
plus our own checks of speed, security and accessibility.
How did you find it?
- Listed
- August 2026
About Lucid
Lucid watches a language model think. It reads the concepts a model holds before it speaks, layer by layer, using Anthropic
No reviews yetWrite the first review
Ratings & reviews
No ratings yet
Be the first to review it. Honest takes are welcome, good or bad.
Share this listing on XOur review
Lucid lets you type a prompt and then watches which concepts a small open language model is holding internally as it works toward an answer, shown layer by layer. The reading is presented as a set of cards, a nod to the Zener cards used in early telepathy tests, so the experiment feels like a parlour game and an instrument at once. Some concepts surface in the model's workspace but never make it into the spoken reply, which is the odd part worth staring at. It suits anyone curious about interpretability who wants to poke at the idea without installing anything. The honest limitation is that the readings only run on models around half a billion to three billion parameters, so the window into the mind is a small and simple one.
- Originality
- 84/100
- Usefulness
- 62/100
Written by an AI model after comparing this app with similar listings.
What we checked
Whether it uses HTTPS and standard browser protections.
Tech Stack
Features
Score up 8 points since August. Last checked November 22, 2026.
More like this
Made this? Claim the listing to edit it, for your site, or promote it.