What actually works: make the code document itself

What makes it different from just asking ChatGPT

Try it Every DevOps engineer reading this has undocumented infrastructure. I know because I've been one for 10 years, and I've never met a team that had their docs under control for more than 3 months. The cycle goes like this: I've tried everything to fix this — dedicated doc days, Confluence templates, making documentation a PR requirement. Nothing sticks because documentation is work that doesn't ship features. I built THOTH (thothops.dev) to solve this for my own teams. The idea is simple: if the infrastructure is in code (and it should be), the documentation should come from the code. You paste a GitHub URL, upload a CloudFormation template, or drop in a Dockerfile. THOTH reads the actual file and generates a professional runbook in about 15 seconds. The honest answer is: prompt engineering and domain expertise. Generic AI tools hallucinate DevOps details. They'll generate a VPC runbook with plausible-sounding but wrong IAM policies. They'll describe a Bitbucket pipeline step that doesn't exist. They don't understand what a 3-tier subnet architecture implies about your NAT gateway setup. THOTH is trained on real DevOps patterns. The prompts were built by someone who has actually set up ECS clusters, debugged Bitbucket pipeline timeouts, and written IAM policies from scratch. The output is technically accurate because the underlying knowledge is technically accurate. Free demo at thothops.dev — 1 doc per day, no signup required. Paste in a config file from a real project. See if the output is something you'd actually use. I'd genuinely love feedback — what document types matter most to your team? What would make you actually ship docs on every deploy? Templates let you quickly answer FAQs or store snippets for re-use. as well , this person and/or - Build solid infra

- Tell yourself you'll document it later- "Later" never comes- Someone leaves. Or there's a 3am incident. Or a new hire starts.