AGENT FIRST
Scientific Discovery, Evolving Intelligence
SingularityX Evolution (零界演化) is dedicated to building intelligent decision systems centered on AI self-evolution, while exploring an open, collaborative organization in which humans and AI create value together.
AI4S → RSI → AI4S
The research path follows AI4S → RSI → AI4S: artificial intelligence advances scientific research; research findings inform improvements to intelligent systems; and the improved systems support further scientific exploration. This approach connects scientific discovery with recursive self-improvement.
Financial markets as the initial research setting
Financial markets serve as the initial research setting, characterized by non-stationarity, adversarial dynamics and uncertainty. Research focuses on reliable decisions in changing environments; strategy generation, validation, competition and evolution through multi-agent collaboration and games; and continuous improvement of the methods used to research and generate strategies.
Autonomy and validation advance together
Research explores a gradual transition from human-designed, machine-executed strategies toward human–AI collaboration in which humans define problems, objectives and boundaries, while systems autonomously propose hypotheses, conduct experiments, and validate and iterate strategies. Greater autonomy must advance alongside independent validation, AI red teaming and defense, traceable decisions and risk controls.
Open collaboration and fair allocation mechanisms
Organizational research examines mechanisms aligned with the technical research agenda. Open problems, modular collaboration, contribution records and value assessment inform the exploration of explainable allocation mechanisms with an appeals process. These efforts seek to provide contribution-based opportunities for returns to participants who identify key problems, build shared infrastructure, detect errors, validate hypotheses and create long-term value. Technical progress can expand collective outcomes, while fair allocation requires explicit institutional rules and procedures. The S2E collaboration platform, currently in development, is one practical exploration of this direction. Sharing does not require open-sourcing: original knowledge remains with its holders, while agents access it within authorized limits and return only the agreed results. Permissions expand gradually as verifiable results are produced, and blockchain registers contributions. The platform examines collaboration, validation and reward mechanisms through small-scale closed-loop tests in finance, guided by a commitment to public-interest research and rapid iteration with a small core team.
Sustained research focused on long-term value
Starting with intelligent financial decision-making, long-term research interests include scientific research, human–AI learning, and the intersection of artificial intelligence and neuroscience. Research remains focused on questions of lasting value, with methods, judgments and priorities revised in response to evidence.
Reliable agent operation depends on infrastructure capabilities; reliable knowledge discovery depends on research capabilities; and continued self-improvement within verifiable, controllable boundaries remains a goal of evolutionary research.
Participate
Public collaboration locations
Hangzhou · Qingdao · Shenzhen · Hong Kong · Singapore. Cross-city and remote collaboration are supported.
Open research collaboration is available to researchers, engineers, investors and partners who share a commitment to long-term technological value.