Revocation Watch

Simulated: one made-up token, a fake clock, and made-up delays. Nothing here talks to a real service or agent. The delays do not describe how any company works.

One token is revoked. Who still trusts it?

An issuer revokes a leaked token. Press Play or Step and watch the notice spread to each app.

How apps learn

How one revocation spreads An issuer revokes a token. In feed mode it publishes one entry to a signed feed. A CDN caches the entry and fans it out. Three apps fetch it and drop the grant, one after another. In Today mode each app learns by its own route. Issuer Revocation feedone public list CDN edgecaches, fans out Google GitHub Slack

Alert feed

Revocation feed (live)


          

Every consumer

Consumer How it learns Status Kept trusting Exposure window

The Today delays are made up. They show the shape of the problem, not a measurement.

Add your own consumer

Try this

  1. Keep "Today" selected. Press Step four times. Read the alert feed.
  2. Choose "With a shared feed". Press Play and compare the worst exposure window.
  3. Add a consumer that ignores the feed. Press Play and see who never drops the grant.

Pain point

A broken key or token stays trusted in each app until someone tells that app.

Problem statement

Apps drop a revoked grant at different times. Each gap is a window where the broken key still works. No shared signal tells every app to stop at once.

What sparked it

Per public reports, vendors blocked the eMudhra keys at different times between Oct 17 and Oct 20 2026.

Why this approach

An issuer publishes one signed notice to a shared feed. Every subscriber acts on it, and the page shows who has not.

Hypothesis tested

We believed that a shared revocation feed shortens the time apps keep trusting a broken key. If it holds, the worst exposure window drops in feed mode.

How to try it

  1. Open the Prototype tab. Press Step with "Today" selected.
  2. Select "With a shared feed" and press Play.
  3. Compare the worst exposure window in both modes.
  4. Add a consumer that ignores the feed and press Play again.

What this does not do

Notes

Traces

Prompts used to make this prototype
StepPromptModelToolDate

Cost

Cost to make this prototype
StepModelInput tokensOutput tokensCost (USD)
Total

Measured: the author runs used 196,175 tokens in total. Estimated: the split into input and output tokens (90 percent input, 10 percent output) and the dollar cost, at $2 input and $10 output per million tokens for Sonnet 5.5 (about $0.55). The run reported one total only. The lead session's planning, research, and publishing are not included. The prompts in the Traces section are listed in the order they were given.