Sentient ROMA: Breaking the AGI Structure with Recursive Meta-Agents Sentient (@SentientAGI) has recently surged by nearly 20% pre-market. I suspect it might be poised for a TGE (Test of Things) event, expected within the month. So, I'm just giving it a final push to see if I can accumulate more shares. In the current AI field, while closed-source models (such as GPT-5) offer superior performance, their operational mechanisms are like a "black box," leaving users unable to grasp their inference logic and data sovereignty. Sentient (@SentientAGI) is dedicated to ending this technological monopoly by building GRID, the world's largest open intelligent network. ■ROMA's Recursive and Transparent Orchestration ROMA (Recursive Open Meta-Agent) is a highly disciplined recursive task execution framework and the core of Sentient. ROMA's four core component control flows: ①Atomizer: Automatically determines whether a task is a directly executable "atomic task" or a complex request requiring further breakdown. ②Planner: Reconstructs non-atomic tasks into an ordered sequence of subtasks. ③Executor: Calls LLM and API. The tool performs specific actions. ④Aggregator: Synchronizes and integrates the outputs of subtasks to generate the final solution for the parent task. The release of ROMA marks a milestone for Sentient @SentientAGI open-source AI in the field of complex reasoning: ➡️Performance benchmark ➡️Sovereignty and maintainability ➡️Building an open intelligent ecosystem