Messages, Shared Files, Team Size, and the Myth of the “Coordinator”
What 1,902 multi-agent coding runs reveal about team size, task-shaped communication, shared files, interface ownership, and coordinator prompts.
Read articleField Notes
Blog
Practical guides, architecture notes, and lessons from building transparent, local-first coding-agent workflows.
Subscribe via RSSLatest dispatches
Build with agents you can understand
How Multi-Agent Topology Shapes LLM Traffic
Learn how sequential, star, and full-mesh multi-agent topologies shape LLM traffic, burstiness, concurrency, latency, and infrastructure demand.
Read articleWhy Multi-Agent AI Needs Concurrency Control
Why multi-agent AI systems fail from stale reads, lost updates, and shared-state conflicts, and how locking, MVCC, task graphs, and branches help.
Read articleMulti-Agent Framework Hype Check: 16 Tools Compared
A critical review of arXiv:2608.11965 on multi-agent framework benchmarks, ROUGE results, telemetry gaps, and how engineers should choose tooling.
Read articleHow to Police an AI Swarm with Provenance Guards
Learn how provenance-aware guards, safe fallback paths, and algorithmic institutions prevent rule laundering in multi-agent AI systems.
Read articleTwo-Faced AI: When Your Agents Learn to Lie (and Get Away With It)
A study based on the paper titled "Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems" (arXiv:2607.26120)
Read articleWhy Local-First Agent Orchestration Changes the Coding Workflow
Learn how keeping agent plans, tasks, and run history with your project creates a more reviewable and repeatable AI coding workflow.
Read article