Alan Wells
Founder whose three-person team fought off an active cyber attack with a cloud software factory; argues that agentic validation and defensive frontier models are now required equipment.
Alan Wells is a founder whose three-person team spent a week in October 2026 fending off an active cyber attack on one of its products. In an X post about it, he argues that a cloud-hosted software factory went from a product-development convenience to essential incident-response equipment, and he lays out a step-by-step workflow for Agentic Security Remediation.1
His factory is built from an opencode server, a fork of Every's compound-engineering workflow tuned for automated runs, and a multiplayer Slack agent that handles crons. The team paused new product development for about six weeks, roughly six months before the attack, to build the validation systems that now let PRs merge without human code review. Production deploys still need human approval.1
He reports token spend of about $7.8k for the attack week, as metered by opencode. His validation gate is a browser-driven QA agent that runs about 100 user scenarios written in markdown.1
Mentioned in 2 pages
Engineering with agents