F2Pool Co-founder Mao Shixing: Decision-making with Three Specialized AI Agents
F2Pool co-founder Mao Shixing stated in an interview on August 25 that daily information is almost entirely processed by artificial intelligence, which operates three specialized agents: one provides rational advice, another explores inner feelings, and the third identifies blind spots by comparing with previous day’s actions.
He mentioned that one can see a summary of the previous day's materials as soon as they open their eyes. The agents can suggest that a historical chapter is similar to the current state and recommend a fifteen-minute reading. There is also a note-taking system that crossbreeds past focal points based on evolutionary theory, with the survival of propositions determined by his attention; daily reports and long articles are processed by the model before being delivered. Regarding the potential lag in real-time information, he stated that the most important information will come proactively. The observation system and agents will also remind him of the existence of death anxiety.
This is a description of a personal workflow, not a product launch or revenue audit. The three agents, if using the same model but different prompts, are role-playing rather than having independent consciousness. Crypto practitioners outsource decision-making to models, and the training corpus of the models does not automatically include their positions and competitor data.
The filtering layer increases the signal-to-noise ratio and leaves the "what to see" to the prompt authors. The information that comes proactively may still be algorithmically fed.
In market mechanisms, the buyer is the decision-maker looking to compress reading time, while the seller charges for models and note-taking tools based on tokens and subscriptions. The driving force is personal efficiency experiments, not price events. Funding shifts from executive time to interface fees. Beneficiaries are platforms that can access multi-role prompts, while those under pressure are traditional media competing for attention with daily report titles.
Source: Public Information
ABAB AI Insight
The founder of the mining pool splits the brain into three prompts: rationality, feelings, and review. This is akin to bringing the red team, blue team, and internal audit of a board meeting into a chat window. Attention as 'God' determines the survival of propositions, acknowledging that the power of selection is scarcer than the power of information. Assigning death anxiety to the agents is a way to productize psychological counseling. The backgrounds of F2Pool and Matrixport imply that their information flow is inherently market and computationally biased, and even a neutral model will be shaped by the feeding format.
The capital path is personal computing rental. There are no new public company subjects, just changes to the founder's calendar. The motivation is to maintain judgment bandwidth amidst information overload; the strategy is to let the model crossbreed first and then read only one article, avoiding becoming the first filter. Real-time relevance yields to 'important information will come,' which only holds true for non-minute-level decisions in trading windows.
Analogous objects include Ray Dalio's principles diary, military post-war reviews, and integrating psychological counselors with analysts into the same interface. The crypto circle is transitioning from K-line group chats to private AI agents, marking a phase where tools evolve from retrieval to personality division.
This is a technological replacement. What is replaced are assistants, analysts, and some introspection. The mechanism is: changing system prompts in the same model changes roles, with the cost being multiple dialogues rather than hiring more people. Once the replacement is complete, errors will appear in three tones simultaneously, sounding like consensus. True risk control remains position-based, not the wording of the third agent.
ABAB News · Cognitive Laws
- When the three agents speak in unison, one should first ask if they are using the same set of weights.
- Attention as God makes the power of selection more valuable than the power of information.
- The most important information will come, provided you haven't handed the front door to recommendation algorithms.