Take a look at Peter Thiel's only bet in the field of open source artificial intelligence is on the business progress of Sentient, an AI project. Peter Thiel founded Silicon Valley's top venture capital firm, Founders Fund, which has invested in a few well-known large model AI companies such as eepMind, OpenAI, Scale AI, Anduril, and Cognition. His investment in the field of open source artificial intelligence is only one - Sentient. We talked about Sentient's overall technical framework over a month ago, and in fact, the entire product and technology stack of its project are already very solid. Now we have launched: 1) GRID: It is an intelligent grid composed of numerous professional small agents working together to share data and computing resources. It has been integrated with over 100 AI models, agents, data sources, and tools, and is currently one of the largest open-source intelligent networks. 2) ROMA: Sentient's open-source multi-agent framework is very active on GitHub (sent agi/ROMA) and has received high praise in the AI community. Exceeding closed source models such as GPT-4o multiple times in terms of complex reasoning, code, legal analysis, and other benchmarks. It has been used by some open source projects and small teams for production level tasks such as code generation, legal/compliance analysis, and multi-source data fusion. 3) SERA: Sentient's semantic embedding and inference proxy, especially the SERA Crypto version designed specifically for Web3 research, can output institutional level on chain analysis reports with traceability within seconds. Currently, it ranks among the top open source proxies. This is Sentient's closest product to real business value at present. It has been widely used in the cryptocurrency community, researchers, and trading teams, and can output on chain analysis reports with complete traceability within seconds (project research, sentiment analysis, whale tracking, token fundamentals, etc.). In terms of encrypted intelligence speed and accuracy, it significantly exceeds the universal versions of Perplexity/ChatGPT/Gemini, especially when dealing with real-time on chain+social+news mixed data. 4) Dobby series models: open-source 8B to 70B parameter models, multi version iterations (Unhinged/Leaded), strong performance and commercial applicability. Sentient's technology stack has achieved excellent results in many key technology evaluations: 1) On the FRAMES benchmark (multi-step inference), ROMA achieved an accuracy of 81.7%, significantly surpassing its competitors; 2) On the SEAL-0 benchmark (complex multi-source inference), ROMA scored 45.6%, more than twice as high as Gemini 2.5 Pro (19.8%), and also defeated Perplexity and some OpenAI/Anthropic models; 3) On the FRAMES benchmark, ODS indeed outperforms Perplexity and GPT-4o (with higher accuracy), and achieves the best open source accuracy on SimpleQA (OpenAI benchmark); 4) SERA (Semantic Embeddings Reasoning Agent, especially SERA Crypto version) is Sentient's inference agent, which does rank first on the DMind benchmark (open-source agent leaderboard) in the open-source category, It can be said that Sentient products have moved from the laboratory stage to the real user desktop stage, belonging to the project of "extremely hard technology, truly usable product, and truly used by the community/developers". More noteworthy is the token economics of the project, where SENT is used for proxies, models, data services, and other artifacts driven products throughout the entire ecosystem. Artifacts can also be paid using SENT to achieve a composable on chain value stream between services. SENT allocation is divided into five categories: community initiatives and airdrops, ecosystems and R&D, teams, investors, and public offerings. Especially in the team and investor sections, after TGE, the team token fund pool will be locked for one year, and then the funds will be released linearly within 6 years, with a one-year freeze period. The investor token will be locked for one year and linearly unlocked within four years. This design can be said to be the most long-term oriented and anti runaway token mechanism in the history of Crypto. The team and VC have little short-term cash out space, and can only make profits if the project and product achieve long-term growth From these perspectives, Sentient is currently the only project that combines excellent technology delivery (GRID+ROMA+OML), community friendly tokenomics, and a truly anti-monopoly open-source AGI narrative. It's not just another AI project, but an attempt to redefine the underlying protocols of who owns AGI and who benefits from it. Its development space is highly imaginative, benefiting from TGE catalysis in the short term. If open-source AGI becomes mainstream in the long term, the spatial prospects are promising, but this also requires time and actual progress of the project to be realized.