I had used Prisma before, but the pieces were fuzzy. I went back through client, adapter, migrations, and using it with Neon.

What I re-learned

• Prisma Client — generated, typed API from schema.prisma. You query models, not raw SQL, unless you opt in.
• Adapter — the driver layer. Neon needs the Prisma adapter for its serverless/HTTP driver instead of a long-lived TCP connection.
• Migrate vs db push — migrate is for real apps (history, deploys). db push is fine while hacking locally.
• Neon — serverless Postgres. Pair it with the adapter + connection pooling so you do not open a new connection on every request.

Mental model

schema.prisma → generate Client → adapter talks to Neon → migrate applies schema changes.

Next

Stand up a tiny Next.js + Prisma + Neon experiment and actually run a migrate, not only db push.