Chapter 14. The Multiagent Reliability Architecture
The single-agent fantasy is the new monolith. It fails the same way, for the same reasons. An agent capable of doing everything is an agent that cannot be reasoned about, cannot be governed at scale, cannot be calibrated by action class, cannot be quarantined when it goes wrong, and cannot be promoted on the evidence of one capability without simultaneously promoting all the others. The argument for multiagent reliability is the same argument that retired the monolith two decades ago: specialization is what makes complex systems tractable, and tractability is the precondition for reliability at production scale.
Many chapters of this book have used the term the agent in service of installing concepts cleanly. The Capacity Agent in Chapter 5, the Test Agent in Chapter 6, the Chaos Agent in Chapter 7, the Observability Agent and the PRS-producing pipeline in Chapter 8, the agents producing Why Traces in Chapter 9, the agents being governed in Chapter 11, the agents being simulated in Chapter 12, and the ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access