Rehearsal: agentic interview playback

Let interviewers see how candidates actually work with AI coding agents

Problem

AI coding agents are changing technical interviews, but interviewers can't see how candidates actually use tools like Claude Code or Codex — which prompts they write, how they recover from errors, whether the AI wrote everything. Current interview tools assume the old whiteboard/LeetCode paradigm, and candidates don't want to hand over personal accounts.

Solution

A lightweight session capture and replay panel: candidates work in their own environment while the tool records the prompt-and-response timeline and file changes; interviewers replay it in the browser with annotations, follow along live, and get a structured evaluation summary afterward.

Users

Startup interviewers and hiring teams running AI-assisted technical interviews

MVP

Local CLI on the candidate side capturing prompts and file changes; browser replay timeline; interviewer annotations and scoring; one-page evaluation summary; macOS first

Revenue

Per-interview pricing ($5–10 each) or a team monthly plan; free for candidates

Why now

AI coding is now mainstream in hiring, but interview methodology lags behind. A solo dev can ship a lightweight protocol tool without replacing the ATS.