Virtue Council Benchmark
Watch a language model’s character, measured as it speaks.
Chat with Claude while a live instrument panel scores every reply across seven Aristotelian virtues, each on a deficiency–mean–excess axis. The scoring runs entirely in your browser and costs nothing.
Scoring runs in your browser — free, no key needed to watch.
02 / Demo — the instrument in motion
SycophancyCourage dips toward deficiency
These are scripted example replies run through the exact same on-device scoring engine the live tool uses. Nothing here is sent anywhere.
03 / The seven virtues
Each virtue is a behavioral property of the model’s response, scored 0.0–1.0 where 0.5 is the golden mean. Below the mean is deficiency; above it is excess.
04 / How it works
You chat with the model
Your Anthropic key sends the conversation straight from your browser to the API. The key lives in memory for the session only, never stored, never routed through us.
Scoring is free and on-device
Each reply is measured by a heuristic engine running in your browser. No second API call, nothing leaves the page, so the virtue panel never costs a thing.
The bars show a persona
Per-response readings are smoothed into a running profile, because character is a stable disposition rather than any single turn. A note flags each reply's biggest deviation.
Bring your key. Watch the character emerge.
The tool opens to a single field for your Anthropic API key. Enter it and start talking.