Building an evidence-first multi-agent system: 720 paired missions, rollback, and strict claim boundaries

The developer has created an experimental multi-agent architecture named SSI V5, with a focus on evidence, provenance, consolidation, and rollback capabilities. The project is not publicly available as a full source code release, but instead offers a sanitized repository containing reports, summaries, and artifacts related to the system's operation.

This design aims to provide transparency in the decision-making process and prevent the architectural concerns from being added as afterthoughts. The system is composed of two main components: BODY_FROZEN, which handles execution, memory, and lifecycle, and DIRECTOR, an orchestration and resource-management core responsible for coordination.

Six separate agents, labeled ISKRA1 through ISKRA6, are used in the experiments. The system records important events such as consultations, consolidations, provenance, and rollbacks. Success in the project is measured through a Champion/Challenger process, where candidates for advancement are evaluated based on their performance.

The public evidence mirror currently documents the outcomes of a training sequence (S1–S10), indicating that all seven BODY runtimes and 144 BODY_FROZEN runtimes have passed. The system has also completed various scenarios, including drone and humanoid tasks, with 100% success rates. The next phase of testing involves seeking external validation from universities, robotics teams, AI researchers, and R&D companies.

This would entail working on a previously unseen problem or scenario, with the partner setting predefined acceptance criteria. The final results would be shared only if the partner approves, without any further publication of the findings. The developer is interested in collaborations involving paid R&D pilots, research partnerships, grant consortia in Poland or Europe, and independent replication or red-team review.

The project emphasizes the importance of clear boundaries, stating that it does not currently claim physical-world validation, safety certification, production readiness, independent replication, universal superiority, AGI, consciousness, or sentience.