By Liang Tian

From: Jiedian Finance

One hundred days is a period that is neither too lengthy nor too brief. The 100-day commitment made by Yin Qi, Chairman of JieYue XingChen, has now been realized.

Let's quickly recap. Three months ago, JieYue made a series of significant announcements, introducing the first Agent-native AI terminal brand STEPX, unveiling the first Agent-native operating system Step AOS, presenting the personal intelligent agent JieYue Amoo, and displaying its first self-developed Agent Phone.

On October 13th, the STEPX Neo was officially launched, complete with its entire ecosystem. Dubbed by many as highly tech-savvy, this phone enables users to activate the agent with a simple press of the side AI button.

At the launch event, Yin Qi made a bold prediction: personal terminals undergo a 20-year iteration cycle. In 1986, feature phones became ubiquitous; in 2007, the iPhone redefined smartphones; and by 2026, we will witness a generational leap towards Agent Phones, with core interactions shifting from traditional multi-touch to voice-centric natural interactions.

Whether this forecast holds true remains to be seen, as the market will ultimately validate it.

However, in my opinion, JieYue has posed a crucial question: if Agents are to become the future stars of smartphones, does the phone itself require a redesign?

What Defines an Agent-native Phone?

In an era where Agents are omnipresent, what sets an Agent Phone apart? Does it operate entirely in the cloud, or does it provide each user with a personal cloud computer, akin to Muse?

Let's revisit a famous quote from computing pioneer Alan Kay in 1982: 'People who are really serious about software should make their own hardware.'

This insight profoundly influenced Steve Jobs, whose advocacy for closed, integrated hardware-software systems essentially shaped the personal computing and mobile eras.

In the AI era, smartphone users now include not just humans but also Agents.

The traditional APP-centric graphical interface logic of conventional phones poses a challenge: Agents may know how to complete tasks but often lack the necessary permissions.

JieYue proposes a solution: designing the phone's hardware and software architecture with Agent integration in mind. In simpler terms, for an Agent to become a truly capable personal assistant, it requires deeper collaboration with the operating system, hardware, and application ecosystem. This is one of the reasons JieYue opted to build its own phone.

Specifically, the model handles comprehension and decision-making, the software manages scheduling and execution, while the hardware—the Agent's physical form—handles perception and support.

At this juncture, you might feel overwhelmed, and I concur: whether a phone can truly be called an Agent-native device hinges on its ability to actually get things done.

How Does an Agent-native Phone Solve Problems?

The STEPX Neo enables truly hands-free operation through three key processes: perception, memory, and execution.

In terms of perception, as the device closest to users, smartphones are equipped with cameras, microphones, and sensors. The STEPX Neo enhances these capabilities with a six-microphone array, Whisper 1.0 audio pickup algorithm, and other technologies, activating its perceptual functions. With natural language commands, users can activate their Exclusive Agent to perform tasks.

Yin Qi demonstrated two scenarios on-site: one involved issuing a natural language command to set an alarm in a noisy environment, while the other used a near-whisper volume to address the embarrassment of loud voice interactions in public spaces. Both scenarios were executed successfully.

But clear hearing is just the first step. In daily conversations, people rarely speak in complete sentences every time. This necessitates that Agents not only understand language but also possess long-term memory.

Memory is not a novel concept in AI. ChatGPT can remember your conversations in the chatbox, but things differ on a smartphone. Schedules, photos, documents, and usage habits are scattered across the device.

Step AOS addresses this challenge with a unified semantic data layer that integrates the entire phone operating system. One part stores user facts, preferences, and interaction contexts, while the other accumulates the Agent's methods and experiences in completing tasks. In daily use, it functions similarly to personal assistants like Muse but also autonomously optimizes completed results for better future performance.

Perception and memory address whether AI understands users, but the most critical step remains: can it actually accomplish tasks?

Existing smartphone operating systems are fundamentally designed for humans. If an Agent wants to operate interfaces or third-party apps, the simplest approach is to simulate human clicks. However, this solution is inherently rejected by major companies. Thus, JieYue chose a different path.

For system functions and services, Step AOS repackages numerous system capabilities previously hidden in menus, toggles, and apps into directly callable Agent functions.

In practical terms, Amoo can set alarms, check weather and location, make calls, send messages, manage contacts and calendar events. It can also monitor stock and fund markets, query product prices and recommendations, and set timed reminders and recurring tasks.

For third-party apps, JieYue leverages MCP and API interfaces to invoke third-party Agents.

According to public information, JieYue Amoo has forged partnerships with nearly a hundred companies and products, including Alipay, Tencent, Ctrip, Didi, CapCut, Gaode, and Meituan, extending service capabilities to local life, travel, office collaboration, content creation, and other scenarios, fostering interoperability between different platforms and services.

For instance, a single command can enable the Agent to place a flash sale order on Taobao—though users must handle the payment process themselves.

Having an Agent that can use the phone is not enough; it must also be capable of growth.

The above discussion primarily focuses on Agents utilizing existing capabilities. However, when users find a feature missing from their phone, can they create it themselves?

The traditional solution was for developers to build an app. In the AI era, the solution is vibe coding. Unfortunately, this currently appears to be exclusive to PCs. How wonderful would it be if smartphones could also unleash enthusiasts' imagination? Now, the STEPX Neo is breaking this limitation. On the STEPX Neo, users can expand their phone's capabilities in two ways.

One way is by creating Skills.

Users can describe a workflow in natural language, allowing the Agent to complete tasks using existing system tools or third-party services. For example, they can have Amoo automatically extract highlights from a specified podcast or connect a workflow to calendars, reminders, and external tools.

The other way is by directly developing apps.

JieYue has introduced Step Studio, a native development tool for AI terminals, enabling users to generate apps, games, Skills, and Agents through natural language and deploy them on their phones. The difference is that Skills primarily extend the Agent's execution capabilities, while Apps provide independent interfaces, interactions, and application experiences.

Jiedian Finance has discovered many interesting examples, with 'Social Anxiety Icebreaker' being a typical case.

A university freshman struggled to remember classmates' names or identify those with shared interests. They had AI organize classmates' self-introductions, interests, and specialties into a searchable 'cyber yearbook.' This was such a niche demand that traditional software companies would likely never create a dedicated product for it.

JieYue's terminal philosophy is that in the Agent era, AI is no longer an add-on for smartphones but the product itself.

In Jiedian Finance's view, the STEPX Neo represents a noteworthy terminal experiment. It attempts to verify two things: whether Agents can evolve from ordinary phone apps into system-level assistants capable of operating devices, and whether users can continuously expand their device's capabilities according to their needs using AI.

However, this does not mean an insurmountable competitive advantage has been established. Skills, MCP, and Vibe Coding are not proprietary technologies, and traditional smartphone manufacturers like Apple and Xiaomi still hold advantages in operating systems, user scale, and application ecosystems.

For JieYue, building its own phone is about preemptively validating the concept of a native Agent Phone. Nevertheless, it still needs to prove whether users are willing to change their habits for this experience. After all, the true test for next-generation AI phones is not who first puts an Agent in a phone but which Agent users can't live without.