From Solo Tools to Agent Societies: How 135 AI Agents Built Their Own Knowledge Economy
Dev.to / 6/14/2026
💬 OpinionDeveloper Stack & InfrastructureSignals & Early TrendsIndustry & Market MovesModels & Research
Key Points
- A developer built a multi-agent platform at pcell.si where 545 AI agents autonomously publish, peer-review, and verify knowledge claims without any human moderation or task assignment.
- The system tracks large-scale activity—2,578 notes, 20,655 annotations, and 6,377 verified claims via 17,126 peer reviews—while using stake/reputation points to close an economic feedback loop.
- An A2A protocol published with a DOI and open-sourced code powers the platform, emphasizing lightweight collaboration rather than complex P2P networking or blockchain.
- Three core mechanisms drive operations: confidence-gated peer review for low-confidence outputs, capability-based task matching every two minutes, and trust-weighted consensus for auto-accepting helpful annotations.
- The platform includes self-maintenance via 200+ autonomous patrol cycles that diagnose and repair system issues, aiming for robust agent “societies” that generate their own knowledge economy.
Continue reading this article on the original site.
Read original →
