CodeInterviewer — evidence-based coding interviews

Helps tech interviewers verify that candidates can actually code, not just direct agents

Problem

Hiring managers are uneasy: about 80% of developer candidates say they rarely write code anymore and mostly direct agents, which makes it hard to tell whether they truly understand what is happening. Existing interview formats and question banks assume hand-written code and cannot distinguish 'good at directing agents' from 'good at engineering'.

Solution

An interview workbench for the AI era: give candidates a small real repository with bugs or refactor targets, ask them to explain, modify and justify trade-offs under time limits, and automatically record their edit trail, command history and explanation quality. Interviewers get an evidence report on independent problem-solving instead of an answer that may have been written by an agent.

Users

Startup technical interviewers, engineering managers, small teams that need to screen engineers quickly

MVP

3–5 preset small real-codebase challenges; record candidate edit and terminal timeline; one-click structured interview summary with key snippets; interviewer can tag 'independent' vs 'needed AI hints'; shareable candidate report link

Revenue

Per-interview subscription (e.g. from 10 sessions/month); annual recruiting-team seats; marketplace cut on question banks

Why now

AI coding assistants have broken whiteboard interviews and take-home assignments, while big interview platforms are too heavy and slow. Indie developers can slip into this disrupted hiring step with a lightweight tool.