Orchestration
Multi-Agent Systems
Anthropic and Cognition published opposite headlines about building with many agents, and agreed on the thing that matters most: the real problem is managing context, and the architecture should follow the shape of the task.
5 chapters41 slides27 minUpdated
Your route through the subject
0/41 slides completed · 0/5 chapters · 0 slides read
Start with: Single Agent versus Multi-Agent: The Core Tradeoff
The chapter map
Follow the route or choose a chapter.- 010/7 completeSingle Agent versus Multi-Agent: The Core Tradeoff~4 minReadField checkPlate
- 020/8 completeThe Case For Multi-Agent: Anthropic's Research System~7 minReadPlate
- 030/10 completeThe Case Against: Cognition's "Don't Build Multi-Agents"~6 minRead
- 040/8 completeWhy Multi-Agent Systems Fail: The MAST Taxonomy~5 minRead
- 050/8 completeReconciling the Debate: It Was Never Single versus Multi~5 minRead
Chapters
Chapter 1: Single Agent versus Multi-Agent: The Core Tradeoff
A single agent keeps unified context and stays reliable but runs sequentially. A multi-agent system parallelizes work at roughly fifteen times the token cost.
4 min7 slides: Single Agent versus Multi-Agent: The Core Tradeoff0/7 complete
Chapter 2: The Case For Multi-Agent: Anthropic's Research System
Anthropic's research system uses an orchestrator with parallel worker agents, and it beat a single-agent baseline by about ninety percent on research tasks.
7 min8 slides: The Case For Multi-Agent: Anthropic's Research System0/8 complete
Chapter 3: The Case Against: Cognition's "Don't Build Multi-Agents"
Cognition argues long-running agents need one unified context to contain compounding errors, since split work across subagents causes conflicting decisions.
6 min10 slides: The Case Against: Cognition's "Don't Build Multi-Agents"0/10 complete
Chapter 4: Why Multi-Agent Systems Fail: The MAST Taxonomy
MAST catalogs fourteen failure modes in three categories found in real traces: specification issues, inter-agent misalignment, and task verification failures.
5 min8 slides: Why Multi-Agent Systems Fail: The MAST Taxonomy0/8 complete
Chapter 5: Reconciling the Debate: It Was Never Single versus Multi
Anthropic and Cognition are both right for their tasks: context engineering decides reliability, and read-heavy versus write-heavy work sets the architecture.
5 min8 slides: Reconciling the Debate: It Was Never Single versus Multi0/8 complete