Shared Memory and Blackboard Patterns

Two approaches coexist in 2026 multi-agent systems: the **message pool** (everyone sees everyone's messages, as in AutoGen GroupChat or MetaGPT) and the **blackboard with subscription** (agents subscribe to relevant events, as in Context-Aware MCP or the Matrix framework). Both are the only stateful part of a multi-agent system — which means both are where the interesting bugs live. The reference failure mode is **memory poisoning**: one agent hallucinates a "fact," other agents treat it as verified, and accuracy decays gradually in a way that is much harder to debug than an immediate crash. This lesson builds both structures from stdlib, injects a poisoning attack, and shows the three mitigations that actually work in production.

Learn + Build 1.3 hr Python (stdlib, `threading`)

Prerequisites: Phase 16 · 04 (Primitive Model), Phase 16 · 09 (Parallel Swarm Networks)

#Problem

Multi-agent systems need a place for agents to share facts. A literal option is "pass everything in messages" — but that reinvents shared state with extra copying. Another is "give everyone a global log" — but global logs grow unbounded and poison easily. A third is "project a view per agent" — scalable but schema-heavy.

When one of the agents hallucinates and writes the hallucination to shared state, every downstream agent that reads that state adopts the hallucination as fact. By the time the human notices, the reasoning chain is five steps deep and the root cause is the third message ever written. Debugging multi-agent accuracy decay is harder than debugging a crash.

This is memory poisoning. It is the second-most-documented failure family in the MAST taxonomy (Cemri et al., arXiv:2503.13657) and it is structural: any shared-memory design without provenance and an unwritable verifier will exhibit it eventually.

#Problem

Keep going. The rest of this lesson is one step away.