Sol0mon – Exploring multi-agent AI engineering and decision verification

A developer has released Sol0mon, an experimental multi-agent system that coordinates independent AI models to research problems, surface disagreements, debate competing approaches, verify claims and optionally carry decisions into implementation and testing, and is seeking feedback on whether the approach is worth pursuing. The developer said the MVP is complete but that they ran out of credits before testing the credibility of verified results against a frontier model, and that the design is based on a research paper finding multi-model debate produces better results — a claim the developer says now needs retesting because frontier models have progressed so rapidly. The project uses browser access as a low-cost alternative to APIs, which the developer says does not break terms of service only when used for individual purposes, with API usage possibly added for enterprise use.

I’ve been building Sol0mon, an experimental system that coordinates independent AI models to research problems, identify disagreements, debate competing approaches, verify claims and carry decisions into optional implementation and testing. The goal is to explore whether structured disagreement and verification can make AI-assisted engineering more reliable than relying on a single model. The MVP is developed, and the project is available for inspection: I’d appreciate feedback from people working on AI agents, orchestration, evaluation and developer tooling. In particular: Is this a problem worth solving, and what would you want to see demonstrated before considering using a system like this? Those are interesting feedbacks, thank you I will try to get it fixed. I had ran out if credits just before testing the credibility of the verified results against a frontier model. I was able to complete the mvp. So I wanted feed back before investing more in something that might not be viable. The browser provides low cost alternative to apis but it doesn’t break the tos only if used for individual purposes. I might add the api usage for enterprise use if needed. This is actually based on a research paper that said multi model debate creates better results. But frontier models have progressed so rapidly, it needs to be tested.