Well, I was reminded why I wanted to make Gaffer in the first place a couple days ago. I had to spend a few hours cleaning up an S3 bucket of playwright test artifacts that a cron somehow missed. I took a look at the code base and remembered how much we over-engineered it; full k8s cluster, NestJS backend, Go upload service, NATS, Postgres and even a Traefik reverse proxy. Classic "build an enterprise architecture for a side project" mistake. The infrastructure cost was bearable, but the cognitive load was insane, and crushed any momentum I had.
Well, I decided to start working on a V2 and to rededicate myself to building in public. I've paired down the architecture to Nuxt and Nuxthub (soon to be Vercel, RIP?) deployed on Cloudflare. Let me tell you, this has been awesome.
Sometimes shuttering a project and rebuilding it with fresh eyes can actually pay off...but we'll see.