Living Onboarding Docs for Fast-Growing Codebases

Engineering teams' onboarding docs are wrong the week after they're written. New hires burn senior time with archaeology questions; AI coding assistants help write code but don't teach architecture intent. The product: a CI-integrated service that maintains living architecture docs — system maps, 'why is it built this way' records, onboarding paths — regenerated and diff-reviewed on merge.

Honest assessment: this space is active (Swimm pivoted here, Mintlify expands, GitHub Copilot workspace explains code on demand) and 'docs generated from code' misses the tacit-knowledge layer that makes docs valuable. The defensible wedge is decision-capture: lightweight ADR tooling that interviews the merging engineer ('why this approach?') and compiles the answers into onboarding narratives.

Onboarding engineers takes 3–6 months partly because architectural intent lives in departed employees' heads. Existing doc tools rot because updating them is manual and unrewarded.

Engineering leads at 20–200-engineer companies with hiring velocity; platform teams accountable for onboarding time.

AI made code-level explanation free, which paradoxically raises the value of intent-level documentation (the part AI can't infer). Post-2024 hiring rebounds put onboarding time back on eng-leadership dashboards.

Swimm, Mintlify, Notion+discipline, Copilot's explain features. Crowded middle; the interview-driven decision-capture angle is the only thin spot. Buyers are skeptical from prior doc-tool disappointments.