About
Cofounder at CybeDefend. Lille, France. More than half of new code in serious engineering teams is now written by AI agents. Claude Code ships 5,000 lines before lunch. A senior reviewer can audit maybe 200 on a good day. The math doesn't work. The bugs that slip through aren't CVEs. They're business-logic rules that live in your team's head. Things like "refunds above €500 need a finance manager" or "this query has to filter by tenant". Scanners can't see those. So we stopped trying to fix scanning. We put security inside the coding agent's loop, with rules mined from your own codebase and injected before the first line of code gets written. Launching our MCP on May 12, 2026. Say hi if you're building in AppSec, AI dev tools, or agent infra.
Links
Badges
