Mastering AI Agent Architectures: Decentralized Styles: Peer-to-Peer, Actors, Markets, and Swarms (Mastering AI Agent Architectures — The Styles Series) - Tapa blanda

Sinopsis

No master to lose. What does that cost, and what does it buy?

Decentralized agent systems replace the coordinator with peers that negotiate, gossip, bid and swarm. This book measures twelve such styles on the Gauntlet benchmark and finds where redundancy pays and where it only adds a bill.

• Peer-to-peer negotiation and gossip for shared state
• The actor model as an agent runtime — the cheapest reliable style in the series
• Contract nets and auctions — verification without a master
• Markets — including an honest null result: pricing tasks does not help when agents are alike
• Swarms and stigmergy — the monitoring benchmark that later books try to beat
• Partitions and split-brain, quorums, and normative governance for open systems

Every decentralized style survives the loss of node 0 without a standby. The decision card says when that is worth the redundancy.

Runnable code, frozen results and a lock at tag gauntlet-b3. Book 3 of 7 in Mastering AI Agent Architectures — The Styles Series.

"Sinopsis" puede pertenecer a otra edición de este libro.