01RECURSIVE SELF-IMPROVEMENT

Multi-agentRSIOpen Dimension GateSelf-improving learning agents

RESOLVING00Explore

Beijing Dimension Gate Technology Co., Ltd. builds AI systems that reshape how people learn, express themselves, and turn knowledge into lasting assets. Guided by the unity of knowing and doing, we make understanding, expression, and growth visible. Our product matrix spans meeting intelligence, personalized learning, education OMO/AaaS, and AI consulting platforms, helping individuals record, understand, and improve while helping organizations turn teaching, knowledge, and service into reusable intelligent assets. We want AI to become a continuously evolving companion for every person.

Three questions first

Three Tsinghua alumni and a group of post-00s young people formed an entrepreneurial team.

The bandwidth between people and the world is compressed by language into a narrow stream of eight bits per second. Most AIs evolve with iterative versions, but they don't know you better because of your use, and as much as they use, they don't know the king's mind after all.

With the use of new, more and more to understand your intelligent engine; multiple product lines, one source, common progress and prosperity, iteration more than.

THE LOOP

Although the door is differentThe way is connectedInherited

Meetings, reading, classrooms, consulting, and everyday conversation enter from different products but follow the same path: record, understand, correct, and reuse. A correction in one place ripples through the whole system, changing judgments elsewhere and keeping the engine aligned as one.

01 / 04

Record

The original scene is retained in a searchable and referenced state, and it is no longer scattered in the database, and it is fragmented.

Each correction quietly returns to the next judgment, helping the system become more accurate over time.

Products: the instruments

From the root of technology to the fruit of the product, it runs through the closed loop and has a sound on the ground. Multiple product lines, each of which should be needed, all have fingers and do not fall into the void.

Driving the same set of individual cognitive engines, homogeneous and heterogeneous, with different paths, thus deriving a variety of business models: C-end, B-end, hardware and software, platform sharing.

Who it is for: International students in English-taught programs, venture and sales professionals with dense networks and information flows, and anyone who wants to preserve everyday life.
What it solves: Create a portrait of yourself and document the flash moments in your life
What it changes: Meetings, life, and everyday conversations accumulate into a living personal profile. Every conclusion can be traced to a specific moment and speaker. Corrections you made before become guidance for the next inference.

2026 Q4 Official LaunchLearn more

Who it is for: College students preparing for TOEFL, IELTS, CET-4, or CET-6: clear goals, limited time, and results-first learning.
What it solves: Generic materials, no matter which level you are at. What you want to read is incomprehensible, what you can read is meaningless, and the memorized words are quickly forgotten.
What it changes: The article is generated in real time according to your current level, and the generated word into the word book is repeated at intervals; the portrait of interest will also proactively suggest what to read next, and explain why it is it.

Launch at the end of Q4 2026Learn more

Who it is for: Public schools, private schools, and training institutions, with principals, academic affairs, teachers, and parents in one shared system.
What it solves: AI course summary, AI visualization scheme, AI personalized topic training, AI teacher cloning, AI personalized chemistry companion, constitute an all-link integrated product for education.
What it changes: Students learn, AI records the understanding process, teachers get academic advice, schools see the overall trend, and then reverse personalized learning. A closed loop runs in the same background, the data does not fall to the ground, and directly returns to the classroom.

Deployed edumindx.cn, 2026 Q4 RuntongLearn more

Who it is for: Expert creators with a distinctive point of view, deep communities, and strong fan trust, along with fans willing to pay for personalized outcomes.
What it solves: Knowledge-based payment sells one-time content in the early days, and repurchases are declining; the general model gives you only consensus, neither taste nor consensus.
What it changes: Each creator has a digital doppelgänger that mounts their real-time knowledge graph, undertakes paid personalized consultations with fans, and shares the platform with creators. The marginal cost of personalized delivery approaches zero.

Launch in 2027 Q1Learn more

About us

Tsinghua Gene, AI native team, has received 20 million level of financing.

CEO

Xingyicai

Born in 2001, undergraduate eight-character class at Tsinghua University. At the age of 15, signed a gold medal in the competition Tsinghua University, a former VC industry practitioner. Responsible for product, marketing, commercialization and transformation.

little red book@caijie helps you do it
small universe@Monkey Mountain and Algorithm

CTO

Chen Xupeng

Xupeng Chen

Tsinghua undergraduate four-character class · Dr. NYU EECs, a world-class young scientist, the core results are reproduced by the Turing Award winner, the father of machine learning, and a former TikTok multimodal large model research scientist.

Google Scholar

Chief Architect

Wu Yongxiang

Born in 2003, Tsinghua Computer Undergraduate Word Class, one of the main contributors to GLM-Infra, with a peak of 1400 TPS · 500 million requests.

Eight bits per second is the bandwidth between you and the world.

01
First, leave the part of the speech intact.

02
Let a piece of text grow out of what it looks like.

03
Let the picture be a scene that you can walk into.

04
Finally pieced together along the timeline into a walkable memory universe.

Our vision

The starting point of information equity

Push open the door of dimension so that every soul can be seen holographically and equally

01situation

The most fertile person in the heart may be the most silent

We believe every person carries a high-dimensional inner world: images, sounds, touch, emotions, memory traces, and thoughts not yet formed. Yet the native channel to the outside world is compressed into linear language at eight bits per second. This is one of humanity’s deepest forms of loneliness, and an old kind of information inequality: the richest inner worlds may be the quietest.

02path

The vision of the Dimension Gate is to open those difficult doors

Dimension Gate exists to open the difficult doors. In the era of large language models, AI lets us preserve one-dimensional slices of ourselves: reading unspoken language and turning thought into text, growing text into lifelike images and 3D scenes, and eventually stitching a lifetime of thought fragments into a memory universe one can walk through. Step by step, we move from a narrow version of reality toward one that is wider and closer to the full shape of the self.

03significance

This is an information affirmative action campaign

This is not merely a technical version of mind uploading. It is a movement toward information equity in higher dimensions. We want everyone to have equal access to data, which is why we work on education. We also want every person to be able to deliver their inner universe to the world completely and without distortion, so that the hardest emotions to name, the wildest imagination, and the most private memories can exist, resonate, and be understood as fully as the most eloquent speech.

Careers

Push the door of dimension with us and join the campaign for the promotion of information equality

OPEN ROLES04

News

Dimension Gate’s D102 booth and product display at TCS2026

2026-09-29 · Company News

Dimension Gate Exhibits at TCS2026 as a Technical Support Partner

From international conversations at booth D102 to on-site technical support, Dimension Gate brought its multimodal AI meeting minutes product to TCS2026, connecting product demonstrations with practical service.

Learn more →