Arm Technology's Jishu Community Releases Robot Dog Puppy and Development Tool JishuBuddy

Companion robots for home scenarios are entering a new phase: the child-oriented robot dog below can not only interact in real time, respond to commands such as sit and stand, and perform playful, affectionate gestures, but also has long-term memory and can remember a child's preferences, becoming an embodied playmate integrated into daily life.

This is precisely one of the new achievements of "Jishu Community" released by the Open Source AIOS Alliance initiated by chip IP design and services provider Arm Technology at the West Bund International Technology Consumer Carnival (WTCC 2026) in Shanghai last week: Puppy, the first prototype of JishuFriends.

Small embodied carriers such as robot dogs and smart cars have lower hardware thresholds. Paired with open-source hardware, a unified AIOS foundation, and a development toolchain, they make it easier to get prototypes running and achieve volume deployment, making them a leading track in the industrialization of embodied intelligence.

Based on this, the JishuFriends series focuses on small embodied intelligent devices for the Physical AI field. Its first prototype, Puppy, uses a robot dog as its carrier and integrates capabilities such as environmental perception, voice and visual interaction, and autonomous movement, fully connecting the physical intelligence loop of perception, understanding, decision-making, and execution, enabling AI to perceive the world and respond in real space.

Another achievement unveiled at the same time is JishuBuddy, a new-generation AI hardware development tool for Agentic AI. This tool uses natural language as its interaction entry point and natively adapts to the Arm embedded ecosystem. It can lower the threshold for developing intelligent hardware while supporting developers in rapidly building AI applications; it can also connect simulation environments with physical hardware to enable virtual-real joint debugging, providing an integrated foundation for embodied intelligence developers.

1. Targeting edge AI bottlenecks, the open-source AIOS platform delivers strong results

Today, technological breakthroughs are frequent in fields such as agents and embodied intelligence, and entirely new hardware carriers such as robot dogs and various AI hardware devices are emerging. Edge infrastructure is undergoing a fundamental transformation, and operating systems are accelerating their evolution from traditional resource management tools toward an agent-native form.

In Arm Technology's assessment, AIOS should aim at autonomous task scheduling and interaction between the virtual and physical worlds. Relying on four pillars—upgrading the interaction paradigm, reconstructing the system architecture, breaking through edge AI computing power, and deepening device-cloud collaboration—it should become the new-generation edge infrastructure supporting the deployment of embodied intelligence and agents.

Therefore, Arm Technology is promoting multi-party collaboration among edge chips, open-source hardware, and real-time operating systems to complete the full chain of chips, systems, hardware carriers, and agent task layers for edge AI.

2. Dual product deployment for small embodied devices and agent applications, greatly lowering the development threshold

The open-source AIOS platform JishuFriends mainly targets small embodied intelligent devices, and its first prototype is Puppy, a robot dog with higher approachability for home scenarios.

In addition to basic capabilities such as obstacle avoidance, multi-terrain movement, and automatic following, Puppy has four key capabilities: continuous perception, embodied interaction, family companionship, and personalized growth.

Specifically, it can perceive temperature and humidity, spatial conditions, and the status of people in real time, recognize each family member, and remember household routines and life moments; Puppy can also perform physical interactions, respond to voice commands, proactively give reminders and initiate task suggestions, and accumulate exclusive memories through long-term interaction with family members.

JishuFriends has verified the feasibility of the AIOS platform in carrying long-term memory, scenario understanding, and autonomous tasks in the physical world, and has accumulated a relatively mature development paradigm for small embodied intelligence.

The second major achievement is the open-source AIOS platform JishuBuddy, built for new agent applications. It provides one-stop solutions to development challenges such as AI hardware adaptation, application development, and embodied simulation debugging, while also supporting the accumulation and reuse of debugging experience.

When connecting to AI hardware, JishuBuddy supports two mainstream access channels, SSH and USB/UART serial ports, and can connect to Arm ecosystem hardware such as Linux development boards and MCU sensor control boards, enabling debugging capabilities such as command dispatch, log reading, resource management, and protocol joint debugging, connecting development computers, simulation environments, and real physical hardware into one integrated system.

In the application development stage, the platform adopts conversational development. Through natural language interaction, it can integrate command dispatch, log viewing, and device status monitoring, reducing the cost of embodied intelligence debugging.

In addition, JishuBuddy can work with JishuFriends to complete debugging of simulation environments and real embodied intelligent devices.

JishuBuddy builds a simulation iteration loop for JishuFriends of "issue command, wait for feedback, verify status." Based on multi-dimensional state feedback such as position, posture, and speed, developers continuously adjust robot action sequences and navigation strategies.

At the same time, information such as problems encountered during task execution, run logs, and repair scripts is automatically retained and archived, forming reusable debugging assets.

The early exploration and implementation of this solution benefit from an ecosystem layout built on cross-vendor collaboration.

3. The open-source AIOS platform is activating the edge AI developer ecosystem and driving commercialization

Under this main line, multiple partners in fields such as edge AI chips, open-source hardware, and agent task execution have already implemented their respective roles.

At the computing power level, Rockchip decouples computing power and control through a "main control SoC + AI coprocessor," laying the computational foundation for multi-scenario deployment of small embodied intelligence; in the hardware field, Radxa will help third-party partners quickly build prototypes based on Radxa products, promoting implementation from sample validation to volume delivery; on the agent task side, the open-source community RT-Thread will focus on embodied intelligence projects and work with Arm Technology to promote the combination of agent capabilities and real applications.

With multiple players in place and the two new achievements unveiled, the industrial value of this ecosystem collaboration is becoming increasingly clear through such division of labor and cooperation.

First, as a unified foundation, the open-source AIOS platform can adapt to multiple domestic edge chips while being compatible with mature robot ecosystems and large models, connecting the entire robot development chain and making local inference and local decision-making feasible solutions.

Edge AI covers multiple links including chips, operating systems, hardware, models, and algorithms, and no single vendor can independently cover all links. By linking all parties in ecosystem cooperation, it can share R&D investment, share a standardized foundation, reduce duplicated effort, and jointly accelerate the commercialization of small embodied intelligence.

In the foreseeable future, relying on a unified open-source AIOS foundation, domestic edge chips, open-source hardware, and agent capabilities will continue to integrate, and the developer ecosystem will continue to grow, perhaps giving rise to more lightweight applications and bringing edge AI into real scenarios faster.

The edge AI industry is moving from single-point chip competition toward systemic competition among foundations, ecosystems, and scenarios. This combination of an open-source foundation, standardized carriers, and lower-threshold development tools provides developers with a more general infrastructure.

In the future, more and more edge AI devices of diverse forms will enter reality. They may not all be humanoid robots; they could be robot dogs, inspection carts, or desktop intelligent hardware. And the starting point of these innovations may come from an attempt by an ordinary developer based on this open-source foundation.