Just saw that @SentientAGI won the 'AI Startup of the Year' award at the Minsky Awards. ROMA Search is truly impressive. Multi-step reasoning capability ROMA Search achieved an 81.7% accuracy rate on the FRAMES benchmark, significantly outperforming competitors. This requires AI to perform logical reasoning, information integration, and conclusion derivation in complex scenarios. Traditional search systems can only handle direct queries and struggle with problems requiring multi-step reasoning. ROMA breaks tasks down recursively, ensuring each sub-problem is precisely handled and ultimately aggregated into a complete answer. ⚡️ SEAL-0 benchmark In complex search scenarios involving conflicting information and noisy data, ROMA achieved a 45.6% accuracy rate—more than double Gemini 2.5 Pro's 19.8%. This gap highlights advantages at the architectural level. When search results are contradictory or incomplete, ROMA uses multi-source reasoning to find the true answer, avoiding being misled by incorrect information. Open-source strategy Sentient has open-sourced ROMA on GitHub, essentially building an ecosystem moat. By inviting more developers to contribute, they accelerate technological iteration while establishing standards through community contributions. When the entire industry builds applications on the ROMA framework, Sentient gains control over the next generation of AI search. Technical architecture ROMA isn’t just about stacking models; it achieves parallel processing through intelligent task decomposition. Complex queries are broken into sub-tasks that can be executed simultaneously, with each task assigned to the most suitable specialized agent. This ensures the system maintains high accuracy while handling more complex reasoning chains. The result aggregation process is transparent and traceable, making it easier to debug and optimize. From technical performance to ecosystem strategy, @SentientAGI is setting a new standard for AI search. When search evolves from keyword matching to true intelligent reasoning, the efficiency of information retrieval will be fundamentally transformed.