Talk about Sentient Sentient is one of the few AI projects that is truly engaged in academic research. This time, NeurIPS has accepted four of their papers, which is considered one of the top three conferences in machine learning and has a considerable amount of value. We investigated Sentient's team background and found that their team consists of four major co founders: Pramod Viswanath (Indian American Professor at Princeton University) Sandeep Nailwal (co-founder of Polygon) Himanshu Tyagi (Associate Professor, Indian Institute of Science) Sensys (listed as co-founder in institutional name) And the investment of Indian scholars in AI, especially in the United States, is a double buff. And this architecture solves a major drawback of traditional professor coins, which is their lack of expertise in capital and media operations 。 To be honest, the earliest attention to this project did not come from the academic community, but from Sandeep. Sandeep's energy and resources in Polygon have been evident to everyone over the years, especially his relationship with the exchange. To get back to the point, Sentient wants to do something quite big, it wants to be an AI aggregator. When receiving a user's task, it is first broken down into several subtasks, then routed to the AI that is most suitable for processing it, and finally summarized into one answer. Based on its mechanism, it did indeed outperform Gemini 2.5 in complex reasoning during internal testing (although Gemini is also evolving, and there is currently no independent third-party evaluation), which proves its strength. In addition, a major highlight of their work is that this is open-source AI. After they opened it up, Sentient's code was Stared 4000 times and Forked 600 times. This is also consistent with what they want to do, only by making it open-source can their routers connect to more open-source AI. The banner they want to carry is the leader of open source AI, competing with several closed source AI giants (OpenAI, Gemini, Grok). Therefore, although they cannot currently compete with GPT-5, which is about 40% of the level, they still win in open source. This is also thanks to the unique and magical economic system of Crypto, which (except for exchanges) does not rely on income to make money, and attention has a much higher impact on valuation than income. This may be an important reason driving their willingness to open source.