RelVectors is an AI algorithm developed independently by GeekStation. Its goal is to move AI from a tool that waits for human operation to an autonomous intelligence that keeps working toward a goal.
RelVectors is preparing to enter commercialization. We are looking for research institutions, industry partners and long-term capital willing to define the next generation of artificial intelligence together. Now is an important window to join this technical path.
RelVectors is a relational logic algorithm for artificial intelligence. It studies the relationships among agent, goal, environment, action and result — and how those relationships change — to explore how AI can establish a continuing SELF and complete tasks by using different models.
Transformer underpins a large part of today's language and multimodal models. They compute over representations learned from data, and can generate language, write code, understand images and handle complex problems.
RelVectors studies how an AI agent can keep its goals, relationships, states and experience, and actively organise, choose and call the models, tools and machines it needs.
Not replacing Transformer — driving it. One SELF can call different models; when a model is upgraded or swapped, long-term goals and the process need not break. Here SELF means a technical state of continuing agency and continuity of action, not a proven subjective consciousness.
AI already has strong local capabilities, but most systems still need people to issue instructions, connect tasks, handle exceptions and hold the goal. RelVectors aims to let AI handle long, multi-stage, constantly changing work autonomously — within explicit goals, authorisation and safety constraints.
One request, one capability call, then a person coordinates the next step.
AI organises models and tools, adapts to change, keeps its experience, and asks for human authorisation or intervention when needed.
Robots keep the continuous relation between goal, action and result in a constantly changing environment.
Coordinating perception, goal changes, task switching and execution records across long missions.
Studying continuous task understanding, state association and decision support in complex scenes.
Production, equipment operation and exception handling staying aligned with long-term goals.
These are future research and industrial validation directions. They do not claim that RelVectors has obtained unmanned-system or automotive functional-safety certification, nor that unsupervised commercial deployment already exists in these fields.
GeekStation Business Incubator began with one purpose: incubating technology and empowering founders with frontier tech. In 2011 it started from a tea house by the Olympic Sports Center metro station in Nanjing, and has grown into a professional incubation body with 13 technology business incubators.
RelVectors is looking for partners who can advance this technical path: studying the scientific questions of relational logic and a continuing SELF, designing reproducible comparative experiments, and validating application value in real long-task scenarios.
RelVectors is a research result in relational logic algorithms for artificial intelligence, developed independently in China. The project team judges that it carries hundred-billion-yuan long-term industrial potential, and it is preparing to enter commercialization. RelVectors does not aim to replace Transformer; it explores a SELF that continually maintains agent, goals, relationships and process, organising and calling different models and tools to deliver autonomy for long, multi-stage tasks. Current collaboration directions include joint foundational research, scenario validation in embodied intelligence, drones, autonomous driving and industrial automation, as well as strategic investment and commercialization. Market potential and future applications are development judgements and directions still to be validated.
Collaboration brief copied You can paste it into an email or a collaboration note. The browser blocked copying; please select the text on the page instead.People who join the development of RelVectors get their share of the dividends of the AI era.
