Content & SEO kickoff

Recommendation: MDX in repo, with Velite or Contentlayer for type-safe frontmatter. Skip the headless CMS.

The case against Sanity / Contentful for this project:

• Content team of one (you, eventually one writer). The "non-dev editing UI" advantage doesn't apply.
• ~138 pages authored over 14 weeks, edited rarely after. Not a high-churn editorial newsroom.
• Cost: Sanity starts at $99/mo for a real plan, Contentful at $300/mo. Versus $0 for MDX-in-repo.
• External CMS adds an auth boundary, a webhook, a build trigger, and a separate place to break.
• Existing /blog already inlines content in a TS file. MDX is the natural next step, not a regression.

The case for MDX-in-repo:

• Content lives in Git. Reviewable in PRs, diffable, versioned.
• No vendor lock-in. If we hire a content team later, we migrate to Sanity in a week.
• Velite gives us type-safe frontmatter (title, slug, date, sources) with zod schemas — catches typos at build time.
• Same authoring style as docs (Nextra is also MDX).
• Image co-location: drop a screenshot next to the .mdx file.

Migration plan if you ever change your mind: Velite output schema → Sanity types → one-time content import script. Maybe a week of work.

Recommendation: Nextra.

Why not Docusaurus: it's React-heavy with a more SaaS-template feel, and the default theme doesn't match the editorial Nexpura aesthetic without significant overrides. Two days of theme fighting we don't need.

Why not Mintlify: closed-source, paid, hosts your docs on their domain by default (workarounds exist). The vendor relationship is another moving part.

Why Nextra:

• Built on Next.js, so it lives in the same app. No subdomain, no proxy gymnastics, same deployment pipeline.
• Sidebar nav, search, breadcrumbs, edit-on-GitHub link all out of the box.
• MDX content — same authoring style as the blog. One mental model.
• The default theme is closer to "documentation as a publication" than "documentation as a SaaS landing page."
• We can override the layout for /docs to use our serif headlines and stone-200 borders. Half a day, not a week.

One downside: Nextra v3 is the current major. v4 is in beta. We pick v3, knowing v4 migration is a possible 1–2 day cost in 6 months.

Brief targets Phase 1 in weeks 1–3. Honest estimate at the voice-and-sourcing quality bar you set:

That's 4–5 calendar weeks at a sustainable pace, not the 3 in the brief. Three weeks is achievable if I batch the comparison pages on locked templates and skip Sentry/observability scaffolding I'd normally insist on for production routes. I don't recommend it — the voice bar you set is the long pole, and rushing it produces exactly the AI-feel content you said to avoid.