🗓️ Last updated June 2026. Field names and stability levels track the still evolving OpenTelemetry GenAI semantic conventions; check them against the official spec on GitHub before adoption. There's
pragmaticstack.in18 min read
The irony is "do more" and "know what it's doing" have to scale together. More autonomy means more room for a silent wrong decision. Without observability, you're just shipping confidence you can't verify
The biggest mistake I see founders make is planning infrastructure around user numbers instead of traffic patterns. A job board with 1M monthly users can be surprisingly lightweight if search is optimized and most traffic is read-heavy.
For me, I'd focus early on PostgreSQL indexing, query performance, connection pooling, backups, and basic monitoring before thinking about multiple servers. Most scaling pain arrives from inefficient queries and missing observability long before Hetzner becomes the bottleneck.
Also, if messaging becomes a core feature, that's the area I'd keep an eye on first rather than job listings themselves. That's usually where complexity starts creeping in.
Lara Laurielle
scaling my tech stack daily
This is something I don't see discussed enough. Everyone wants AI agents to do more, but understanding what they're actually doing is just as important