AI Pulse · Under the Radar

A solo builder has documented a setup where one prompt can take a software idea through to a live, HTTPS-protected deployment, without relying on mainstream SaaS tools. The system creates a code repository, writes the application and tests, runs CI, provisions a database, and deploys it. The stack is self-hosted, using Forgejo for code hosting, Coolify for deployment, Hermes and Firecrawl, on an isolated eBay box accessed only through Tailscale.

The interesting part isn’t that every company should copy this exact stack. Most shouldn’t. Running your own infrastructure creates real work around security, backups, monitoring, updates, and failure recovery. But it is a concrete blueprint for firms that care about where their code, customer data, and internal processes live. For a software company, consultancy, or business handling sensitive information, keeping the full build pipeline under your control can reduce dependence on a string of vendor accounts and changing platform rules.

It also shows what an “agentic” workflow looks like when it is connected to real operating systems, not just a chatbot generating snippets of code. The value comes from the chain: request, build, test, deploy, then make the result available safely. That’s the kind of operational visibility and controlled automation we build into an AI command centre.