Chainfeeds Introduction:
From my understanding, the REI framework indeed sounds like it brings us closer to AGI (Artificial General Intelligence). It is a system with memory, capable of learning in each interaction, and intelligently allocating tasks, rather than just predicting the next word based on training content.
Article Source:
https://x.com/blocmatesdotcom/status/1952311167294304678
Article Author:
blocmates.
Perspective:
blocmates.: In recent years, artificial intelligence technology has advanced rapidly, especially with the quick development of large language models (LLMs), giving the illusion that AGI is imminent. From the video of Will Smith eating pasta to nearly indistinguishable realistic images in 2025, we have witnessed unprecedented technological evolution. However, recently, this progress seems to be slowing down. The previous strategies of data stacking, increasing computing power, and expanding parameters are encountering diminishing marginal returns, particularly evident in reasoning capabilities. Despite models becoming larger and more costly, performance has not proportionally improved. This has prompted the industry to rethink: if larger models are no longer the solution, where should we go next? The core issue is that LLMs do not truly think; they merely predict the next word based on context. While this mechanism is sufficient for text generation tasks, they expose clear shortcomings in more complex tasks like planning, reasoning, and learning. REI (Reimagine Intelligence) is a crypto-native AI research team dedicated to rebuilding artificial intelligence from first principles. Their Core framework is a highly modular system that simulates the human brain's structure, decoupling and independently assigning language processing, memory management, and reasoning capabilities. The Bowtie system handles memory and concept formation, differing from traditional LLMs' static context memory. Bowtie has real-time training capabilities, converting user inputs and feedback into permanent cognitive updates. Information is stored in dual-channel encoding, preserving symbolic representations of semantic relationships on one side while capturing high-dimensional context on the other, helping the model continuously build user workflows or thought models. Simultaneously, REI introduces the Reasoning Cluster system, simulating a synthetic brain by allocating advanced cognitive tasks like logical reasoning, causal judgment, and multi-step planning to multiple specialized sub-models for collaborative processing. This design not only improves processing efficiency but also can track each decision path, enhancing transparency and interpretability. Ultimately, Core acts like a conductor, intelligently allocating tasks to the most suitable modules, achieving truly distributed intelligence. Unlike most closed AI companies, REI is not a Silicon Valley venture capital-backed firm, but an open, decentralized research organization aimed at achieving transparent financing and community collaborative innovation through blockchain and internet capital markets. The eponymous token $REI is currently the only crypto asset that provides direct exposure to an AI lab in the public market. Compared to OpenAI, Anthropic, and DeepMind, which are acquired by tech giants or locked in private equity, REI's openness and liquidity offer rare participation opportunities for ordinary investors. Specifically, REI has a total supply of 1 billion tokens, with 54% raising approximately 400,000 USDC through token sales; 30% injected into the Uniswap liquidity pool on Base; 6% to be used on other platforms (decided by community voting). Of the remaining 10%, half is managed by a multi-sig wallet for research grants and experimental projects, while the other half is linearly released through a Sablier contract with a 6-month cliff period. Additionally, the project has reserved $140,000 for future development and recruitment budgets. This design ensures transparency in fund usage while strengthening the alignment of interests between REI and the crypto community. [Original text was in English]
