Context
The current site runs on rented infrastructure that decides how the writing looks and where it goes. This project replaces it with a publishing platform owned end to end. The goal is plain: one place to write, one place to keep roughly sixty existing posts, and a path to push them outward without copy-pasting into three other boxes. It is the foundation the rest of the studio's content sits on.
What was built
So far the platform exists as three specification documents: system architecture, database schema, and a phased nine-stage build sequence. The stack is settled. Node and Express at the core, PostgreSQL with Prisma for data, vanilla JavaScript and GSAP on the surface, Cloudflare R2 for media. Three MCP servers are planned to handle cross-posting: one publisher, one for LinkedIn, one for Substack and its Notes feed. A human review gate sits in front of anything that goes live. Eight architectural decisions stay open. The spec is honest about that.
What was learned
Writing the spec surfaced the real cost: not the code, but the eight decisions that cannot be deferred. Naming them before the build is the work.