A decade in, the microservices promise looks like a tax invoice: tiny repos, noisy alerts, and a big distributed monolith glued together by flaky network calls. The author’s pivot to consolidated “blocks” echoes what small teams already know: start simple, split only when pain demands it, and optimize for throughput, not architecture theater. Fewer deployables mean tighter context for AI tooling, cheaper ops than herding hundreds of pods, and faster upgrades across the stack. I’ll take a modular monolith—modern Java + Quarkus or Rust where it counts—over 300 RPCs to serve a homepage. Keep headcount lean, ship on one box if you can, and scale when it actually hurts.