Articles by tag: ai-coordination

Multi-Agentness as FLP: limits of AI coordination

Breakdown of why multi-agent development is distributed consensus with FLP theorem and Lamport. Practical rules for pipelines: failure detectors, validators. For senior developers.

Physics of LLM-graphs: metrics F and rho

Measuring degradation in multi-agent systems: Fidelity, error correlation, Star and Tree topologies. Analysis for middle/senior dev. Test your swarm on the stand.

Self-organization of LLM agents without roles +14% quality

Experiment: Sequential protocol of LLM agents outperforms coordinator by 14% on Q. 25k tasks, 256 agents, DeepSeek at 1/24 Claude price. Insights for multi-agent systems — read details.