Grab's engineering blog traces their four-year documentation journey from a decentralized Docs-as-Code model (Markdown in individual GitLab repos) to a centralized documentation repository. The decentralized approach solved early problems like scattered docs and lack of version control, but at scale it created fragmented search, inconsistent standards, and governance headaches. The centralized model brought unified search (including Glean-powered AI search), automated linting across all content, consistent templates, and clearer contribution metrics (a commit every ~52 minutes, ~27/day). The post concludes with practical guidance: use decentralized repos when docs are tightly coupled to services, centralize when organization-wide discoverability and governance matter, or adopt a hybrid approach combining both.
