NewJourney Alpha - Tomorrow's NewJourney launches the industry's first cognitive model - AiBoss

What is NewJourney Alpha?

Nextie Alpha is the industry's first cognitive model, launched by Nextie, requiring only 4 bytes of parameters for edge deployment. Led by Li Di, the "father of Xiaoice," the model utilizes an open-source inference model and reinforcement learning to decouple knowledge and cognition, refining its thinking algorithm. In swarm intelligence tasks, the model performs equivalent to GPT-5.4, with a dramatic reduction in computational cost, and supports 24/7 active operation. The model is paired with the Tuanzi multi-agent platform, enabling AI to debate, vote, and collaboratively make decisions like a human expert team.

Main functions of NewCheng Alpha

Cognitive Core CondensationStrip away massive amounts of factual memory, retain the thinking algorithm, and let the model learn how to think.
Swarm intelligence collaborationIt provides unified planning and inference capabilities for the Harness multi-agent system, supporting collaborative thinking among multiple agents.
Local deployment on the device sideThe 4B golden size allows it to run directly on devices such as MacBooks and smart home devices without relying on cloud computing power.
Active continuous operationLow-cost support for 24/7 uninterrupted autonomous planning and execution, shifting from reactive to proactive action.
Five-dimensional cognitive assessmentThe quality of swarm intelligence is quantified based on the completeness of perspectives, the degree to which implicit demands are met, the depth of dialectics, the practicality of implementation, and the interpretability of decisions.

The technical principles of NewCheng Alpha

Decoupling of knowledge and cognitionNextie chose to build upon existing open-source inference models using reinforcement learning. By decoupling knowledge and cognition, it refined the model's cognitive core, focusing its training on generalization and abstraction capabilities, enabling the model to apply its knowledge to different scenarios.
220 Years of Swarm Intelligence Evolution TrainingThe team analyzed human academic papers spanning 220 years from 1800 to 2020, summarizing the evolutionary trajectory of collective intelligence. They deeply dissected the mechanisms by which humans make better decisions through debate, reflection, challenge, and voting, transforming these mechanisms into a machine-learnable cognitive framework to provide theoretical support for cognitive models.
Small parameters + high-quality architectureContinuing the Xiaoice team's approach of small parameters + high-quality architecture. The 4B parameter size has been proven to be the golden ratio, suitable for everything from supporting complex algorithms to being runnable on the edge. Cognitive architecture, rather than parameter size, is considered a long-term competitive advantage.

How to use the new process Alpha

Access PlatformOpen the Tuanzi Multi-Agent Platform beta test address https://mytuanzi.com/ in your browser to experience swarm intelligence collaboration.
Register/LoginDuring the internal testing phase, "power consumption" will be used instead of the computing power points mechanism.
Select a scenarioChoose from preset AI agent combinations such as Sister Group, Research Group, or Ask Miracles, depending on your needs.
Initiate a taskInput your question or requirement, and the system will automatically schedule multiple AI Agents with different expertise to collaborate.
View DecisionsThe goal is to observe the debate, challenges, reflections, and voting processes among multiple agents, and ultimately obtain the optimal solution output after group deliberation.

The core advantages of NewCheng Alpha

Winning big with small stakesModel 4B parameters achieve an equivalent output effect to GPT-5.4 in swarm intelligence tasks, breaking the scaling dilemma that parameters are just.
Costs plummetedIt can run on the edge, eliminating the need for cloud deployment and token billing, reducing computing power costs by an order of magnitude, and changing from burning graphics cards to paying for electricity.
Cognition can be generalizedBy stripping away factual memories and retaining the thinking algorithm, it is possible to transfer the thinking strategies of a single scenario to unrelated fields, achieving cross-domain application of knowledge to new situations.
Transparent and traceable process: Continuing the approach of Little Ice Chain, the agent's thinking process is observable and verifiable, providing a reliable foundation for multi-intelligence collaboration.
Active continuous operationLow power consumption supports 24/7 uninterrupted autonomous planning and execution, enabling the agent to shift from passive response to proactive action.

Comparison of similar products to the New Cheng Alpha

Comparison Dimensions	Nextie (New Journey Alpha)	DIKWP Cognitive Model (Duan Yucong Team)
Core positioning	Lightweight cognitive core engine (4B parameters)	Five-layer cognitive architecture theoretical framework (DIKWPaaS platform)
Cognitive Architecture	Strip away factual memories, retain the thinking algorithm, and strengthen generalization and abstraction abilities.	A five-layered network semantic structure—data → information → knowledge → wisdom → intent—emphasizes the guidance of "intent."
Parameter size	4B (Deployable on the edge)	It is not bound to a specific model and can be superimposed on any large model.
Training methods	Reinforcement learning is based on an open-source inference model to decouple knowledge from cognition.	By formally defining reasoning rules using "semantic mathematics," the cognitive process can be formalized.
Evaluation system	Five-dimensional assessment of swarm intelligence: completeness of perspective, implicit demands, dialectical depth, practicality, and interpretability.	"White-box" assessment of artificial consciousness: analyzing data/information/knowledge/wisdom/intent capabilities layer by layer.
Deployment Form	End-side model + Tuanzi multi-agent platform (productized)	The DIKWPaaS semantic platform (academic/platform-based) emphasizes "semantics as a service".
Human roles	Human participation in multi-agent debates and voting allows for real-time intervention in decision-making.	Human intention, as the highest level, guides the entire cognitive process.
Typical scenarios	Complex business decisions, startup consulting, risk assessment, proactive agents	Active AI, Active Medicine, Artificial Consciousness Research, Semantic Closure Reasoning
Transparency	The thought process is observable and traceable (continuing the small ice chain approach).	Emphasizing "white-box" analysis, each cognitive ability can be independently audited and assessed.
Cost structure	Low power consumption on the device side, continuous operation 7×24	Relying on a cloud-based distributed architecture, the cost is linked to the chosen underlying model.

Application scenarios of NewCheng Alpha

Complex business decisionsCorporate strategy formulation and investment risk assessment utilize a multi-agent perspective to comprehensively identify potential vulnerabilities and external risks.
In-depth research and analysisAcademic research and market surveys, through the cognitive collision of interdisciplinary expert groups, break through the limitations of single-model cognition.
Smart home robotsEdge deployment enables home robots to have continuous autonomous planning capabilities, with controllable computing costs and economic feasibility in daily scenarios.
Entrepreneurship Consulting and GuidanceThe simulation involves multiple agents challenging a business plan from different perspectives, seeking the optimal solution rather than the standard answer.
Content security auditFor example, in the case of Guardian Shrimp, swarm intelligence can accurately identify hidden high-risk vulnerabilities and adjust the output in real time.