Interview Room
One question at a time. Answer it the way you'd say it, and see what was missing.
Speaking uses your browser's speech service, which may send the audio to the browser maker to turn into text.
Grading needs your own Anthropic key. Add it, or watch a demo round first.
Watch a demo round
No key needed. The candidate is invented, and the model's replies were written by hand in the API's format. Everything after the reply is the real grader.
Your key
Paste an Anthropic API key from the Claude Console. It stays in this browser, and it is sent only to api.anthropic.com, with each answer you grade. Usage is billed to your own Anthropic account.
Story bank
Your own stories, kept in this browser. They go to the model with each answer, so a rewrite can use your numbers. A rewrite may only use numbers you wrote here or in the answer.
No stories yet.
Add a story
How it grades
The model reads your answer and copies out which sentences are the Situation, the Task, the Action and the Result. Then this page checks its work and does the counting. Three checks, one point each:
- Structure. All four parts are there, setup is no more than 40% of the words, and the action is at least 30%.
- Evidence. How it ended has a number in it, or a concrete outcome someone could check.
- Length. One to two minutes out loud, at an assumed 150 words a minute.
Three is Ready, two is Close, one or none is Rework. Every quote the model gives is checked against your answer, and a rewrite that uses a number you never wrote is thrown away.
Your stories and your key stay in this browser. Your answer and your stories leave it only when you press Grade, and then only to Anthropic. If you use Speak, your browser's speech service hears the audio. This page has no server, no account and no tracking.
This is practice. It can tell you whether an answer is clear. It can't tell you whether you'll get the job.