Week 4 · Further reading

Many agents, in more depth

The slides covered the main ideas. This page has the details and the evidence, with sources, checked on October 13, 2026. Multi-agent features change monthly: when this page and a product's docs disagree, trust the docs.

  1. Subagents in Claude Code and Codex
  2. Workflows, and who decides
  3. Peer agents and agent teams
  4. The evidence: when it helps
  5. How multi-agent systems fail
  6. Two large experiments, and one for science
  7. A2A
  8. Sandboxes for many agents
  9. Git worktrees

1. Subagents in Claude Code and Codex

Claude Code: subagents · Codex: subagents · OpenAI Agents SDK: multi-agent

2. Workflows, and who decides

Anthropic's "Building effective agents" (December 2024) separates workflows, "LLMs and tools orchestrated through predefined code paths", from agents, where "LLMs dynamically direct their own processes and tool usage", and names five workflow patterns: prompt chaining, routing, parallelization, orchestrator–workers and evaluator–optimizer. Its advice: "find the simplest solution possible, and only increase complexity when needed."

Claude Code's dynamic workflows (May 2026) make the idea concrete: Claude writes a short JavaScript script with agent(), parallel() and pipeline(), and a runtime executes it, so the plan lives in code and intermediate results stay out of Claude's context. Defaults: 16 agents at once, up to 1,000 per run, and no user input mid-run. Anthropic's headline example rewrote the Bun runtime from Zig to Rust, about 750,000 lines with 99.8% of its tests passing, in eleven days (a vendor claim).

Anthropic: building effective agents · Claude Code: workflows · Introducing dynamic workflows

3. Peer agents and agent teams

Claude Code's agent teams (experimental since February 2026; turn on with CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1) are the peer shape as a product: a lead session starts teammates, each a full Claude Code session with its own context (it loads CLAUDE.md, MCP servers and skills, but not the lead's conversation). They share a task list (pending, in progress, done, with dependencies; claiming uses file locks) and a mailbox per agent, messaging each other by name.

Claude Code: agent teams

4. The evidence: when it helps

Anthropic: our multi-agent research system · arXiv 2512.08296 and the Google Research blog · Cognition: don't build multi-agents

5. How multi-agent systems fail

MAST (Cemri et al., NeurIPS 2025) annotated 1,642 traces from 7 open-source frameworks; between 41% and 86.7% of tasks failed. Fourteen failure modes in three groups:

GroupShareFailure modes (share of all failures)
System design~44%step repetition 15.7%, unaware of when to stop 12.4%, disobeying the task 11.8%, losing history 2.8%, disobeying the role 1.5%
Inter-agent misalignment~32%reasoning–action mismatch 13.2%, task derailment 7.4%, not asking for clarification 6.8%, conversation reset 2.2%, ignoring another agent 1.9%, withholding information 0.85%
Task verification~24%incorrect verification 9.1%, no or incomplete verification 8.2%, stopping too early 6.2%

Simple fixes helped only modestly: better role descriptions +9.4%, a verification step +15.6% on one framework.

The echo chamber

The counterpoint: debate helps when agents differ and when verification is grounded in something outside the conversation. Settle disagreements with tools and tests. For research, two agents that read the same sources are not independent reviewers, and agents citing each other can turn one guess into a consensus.

Cemri et al., "Why Do Multi-Agent LLM Systems Fail?"

6. Two large experiments, and one for science

7. A2A

The Agent2Agent protocol (Google, April 2025; version 1.0 in March 2026; at the Linux Foundation's Agentic AI Foundation since August 2026, next to MCP) lets an agent delegate to another agent it doesn't control. An agent publishes an Agent Card at https://{domain}/.well-known/agent-card.json: its name, skills, endpoint and how to authenticate. A client sends a task; the task moves through states (submitted, working, input-required, auth-required, completed, failed, canceled, rejected) and returns messages and artifacts made of parts (text, files, data). Bindings: JSON-RPC, gRPC and HTTP. Over 150 organizations support it; it runs in Azure AI Foundry, Copilot Studio and Amazon Bedrock AgentCore.

A2A specification · Linux Foundation, April 2026

8. Sandboxes for many agents

A container shares the host's kernel; gVisor puts a user-space kernel in between; a microVM gives each sandbox its own kernel behind hardware virtualization. Firecracker microVMs boot in under 125 ms with under 5 MiB of overhead each.

ServiceIsolationNotable
E2BFirecracker microVMpause and resume (files and memory), per-sandbox egress firewall
Modal SandboxesgVisor (a VM runtime optional)snapshots, GPUs, up to 24 h
Daytonadedicated kernel, filesystem and networksnapshots, volumes
Cloudflare Sandbox SDKa microVM per instanceinternet can be switched off; storage mounts
Vercel SandboxFirecracker microVMcredentials injected at the network boundary
Claude Managed Agentsa fresh Linux container per sessionhosted harness; self-hosted sandboxes in beta
OpenAI Agents SDK sandbox agentsa layer over the providers aboveresume from snapshots
Codex clouda container per taskinternet off during the agent phase by default

The pattern from class, for running many agents: a sandbox each, no real key inside (a proxy adds it, limits what can be called and counts it), a time limit and a budget on every run. That is how this class ran the week 1 trials and tested the week 2 harnesses.

Firecracker · each provider's docs, October 2026

9. Git worktrees

A worktree is a second working folder on its own branch, sharing one repository's history: two agents can't overwrite each other's files, and you compare and merge at the end. Worktrees isolate files, not processes, and don't prevent logical conflicts, so still split the work sensibly.

Claude Code: worktrees · Codex: worktrees