Businesses: brand identity and mascots
Give a corporate mascot a body of thought, personality, and way of speaking that fit the brand, so it can introduce the company, its business, and its products through conversation.
Method and boundaries
What is Thinking AI?
Thinking AI is a product and technical method. Building on a general-purpose language model, it organizes domain knowledge, a system of thought, a character’s stance, judgment principles, conversational boundaries, personality, conversational style, and practical skills, so that the AI responds in conversation from a defined body of thought, a clear stance, a distinct style, and explicit boundaries. It is emphatically not a general-purpose AI that complies with every request, flatters the user, or claims to be capable of anything.
Method comparison
How is it different from industry knowledge-base AI?
The table compares the priorities of the two product designs, using an industry knowledge-base AI—one built mainly for document retrieval and question answering—as the reference point. A real product can combine both methods.
| Dimension | Industry knowledge-base AI | Thinking AI |
|---|---|---|
| Content organization | Organizes searchable material around industry documents, facts, terminology, and sources. | Beyond domain knowledge, it organizes a system of thought, a character’s stance, judgment principles, and conversational boundaries, layering source material, its interpretation, and its application in the product. |
| Starting point | Starts from the user’s information need, locates relevant material, and answers the question. | Starts from the specific situation, question, need, or instruction the user describes. It first uses its own thinking and philosophical stance to understand the relationships, choices, and uncertainty involved, then decides how to respond. |
| How the answer is formed | Draws on the material to summarize information, explain concepts, or return search results. | Applies knowledge to the question at hand from its defined perspective and judgment principles, asks follow-up questions when needed, and proposes next steps worth trying. |
| Style of expression | States information clearly around the question; tone and format can be configured for the business context. | Maintains the stance, personality, and conversational style set by the character or brand. How it speaks is part of the product design. |
| Answer boundaries | Mainly indicates whether the material covers the question, where the information comes from, and what it can answer. | In addition to sources, it defines what the character will and will not discuss, when to acknowledge uncertainty, when a question exceeds the character’s abilities, and which decisions to leave to the user. |
| Focus of validation | Checks whether retrieved material, citations, and factual statements match the knowledge base. | In real questions, checks whether it understands the user’s situation, stays within its defined thinking and boundaries, and offers advice that can be used in practice. |
For the specific sources of this thinking and method, read How the Zen Echo (智禅者) Hardware Series Came to Be (in Chinese).
How we build it
No need to retrain a foundation model.
A general-purpose language model serves as the base. Domain knowledge, the character’s thinking, philosophical stance, persona, prompts, product rules and response algorithms, classical source texts, interpretations, and modern product applications are processed in layers and merged. The system then generates a response that fits the character’s style and the product rules. More sources: How the Zen Echo (智禅者) Hardware Series Came to Be (in Chinese).
Possible scenarios
How different customers can use it
Factories: product planning and production delivery
Design character dialogue and practical skills around the target audience and device form, and integrate Thinking AI into companion hardware that can be manufactured.
Designer-toy makers and distributors: character companionship
Design distinct personalities, stories, and conversational styles for designer-toy characters, so users can keep interacting through voice chat, listening to stories, and playing games.
Character IP makers and distributors: character interaction
Embed a character’s thinking, stance, knowledge, and answer principles in the device, so users can talk about the character’s story, its core subject matter, and everyday questions.
FAQ
First, the key boundaries, stated plainly.
What are the benefits of Thinking AI?
It gives a brand or character a consistent body of thought, stance, personality, and style of expression, and lets it answer users’ questions with specific skills. Users can keep interacting with a character that has a defined design, rather than receiving only generic answers.
What can’t Thinking AI do?
It cannot guarantee that every answer is accurate, understand every situation, or make decisions for the user. When a question falls outside the character’s knowledge, design, or capabilities, it should state the uncertainty and the boundary.
How do you verify answer quality?
For now, the founding team continues to test and iterate manually on real questions. A more formal content review system is still being developed.
Does this mean the AI is conscious?
No. Thinking AI is the name of a product and a technical method. It does not mean the model is conscious or has autonomous thoughts.
Put the method into your product context.
Tell us about the character, its appearance, the audience, the use cases, the content boundaries, and the device form. We will use that to discuss a partnership that fits.