I made a AI Code Review tool that knows how your company works
Document-grounded code review with dependency graphs beats generic linters.

Glassdoor alternative with anonymity, but zero network effect—needs critical mass of reviews to compete.
Job seekers, employees evaluating workplace culture, HR analysts
Glassdoor · Blind · Levels.fyi
Live at https://exposetheculture.com
The core idea: reviews focus exclusively on culture — not salary, not interview prep, not benefits. Five rated dimensions: management transparency, work-life boundaries, psychological safety, growth & development, and team collaboration. Plus a binary "would you recommend" and optional free text.
The interesting part is the anonymity architecture:
- Company email is used for one-time verification, then the association is destroyed - Only a hash is stored (to prevent duplicate reviews), never the email itself - Reviews are timing-obfuscated — they don't publish at the exact moment of submission - Metadata is suppressed for companies below a reviewer threshold to prevent inference attacks - The system is designed for genuine data deletion, not just soft deletes
No account required to browse. No employer reputation management tools. No investor pressure to monetize the reviewed-company side.
Stack: Laravel + Blade, PostgreSQL, Redis, Postmark for transactional email.
Would love feedback on the verification/anonymity approach — specifically whether the hash-only model is sufficient or if there's a stronger approach I'm missing.
Document-grounded code review with dependency graphs beats generic linters.
Automated code review loop via agent ping-pong, but Cursor already does multi-turn fixing in context.
Three AI wrappers, but remove.bg, Grammarly, and GitHub Copilot already own these lanes.
Review platform for real estate deals fighting fraud without network effects yet.
Yelp for your friends list, but breaking incumbent network effects is brutal.
Glassdoor for culture values, but job boards with reviews already exist.