Countercurrent
Independent AI-native neo lab · 2026

Against the current.
Toward the inevitable.

Countercurrent is an independent AI-native neo lab exploring the next relationship between humans, intelligence and the physical world.
Countercurrent 是一家独立的 AI-native Neo Lab,探索人、智能与物质世界之间的下一种关系。

Primary wordmark
Countercurrent
Current mark
CC monogram
Cↄ
Research seal
CC
01 / The shift

The interface is no longer the product. The relationship is the product.

The history of computing has been a history of increasing proximity.

计算的发展史,本质上是一部智能不断靠近人的历史。大型机将计算带入机构,个人电脑将计算带上桌面,智能手机将计算带入手掌。下一代计算,将智能带入生活本身。

Intelligence will move across devices, environments and moments; understand context before requesting instruction; form interfaces only when needed; act locally when privacy matters and reach outward when greater intelligence is required.

智能将在设备、空间与时刻之间流动,在请求指令之前理解情境,只在需要时生成界面;在隐私重要时保持本地,在需要更高能力时连接更广阔的智能。

ApplicationsAgents / 智能体
CloudLocal / 本地
CommandsIntent / 意图
Static UIAdaptive UI / 自适应界面
One deviceConstellation / 设备星群
02 / Our thesis

A constellation of specialized, intelligent and interconnected forms.

The post-app era will not be dominated by one universal device. Each form will exist closer to a particular intention, environment or human behavior—simpler on the surface, more capable beneath it.

后应用时代不会由一个万能设备统治。它将以一组专业化、智能化并彼此连接的形态出现。最重要的技术将越来越不可见,最有意义的物件将越来越个人化。

01

Local Intelligence
本地智能

Faster, more private and resilient intelligence operating closer to the person. 在更接近人的位置运行智能。

02

Agent-Native Interaction
智能体原生交互

Systems organized around goals, context and continuous action rather than menus and isolated commands.

03

Adaptive Interfaces
自适应界面

Interfaces that emerge for the moment, modality and intention—and disappear when no longer needed.

04

Distributed Context
分布式情境

Continuity across environments without reducing a person to a single device.

05

Expressive Computing
表达性计算

Technology shaped by identity, culture, material, ritual and desire—not only utility.

06

Trusted Autonomy
可信自主性

Autonomous systems with visible boundaries, human authority and meaningful control.

Some futures need to be built before they can be named.
有些未来,需要先被创造,之后才能被命名。
03 / Signals

Signals of the next computing era.

These are not isolated product announcements. They are evidence of a system-level transition. 这些不是彼此孤立的产品发布,而是一次系统级迁移正在发生的证据。

1 PFLOP

Intelligence moves closer / 智能正在靠近人

Compact systems can provide approximately one petaFLOP of FP4 AI performance in a desktop form factor. The significance is not simply speed. It is proximity.

128 GB

Memory becomes the boundary / 内存成为新边界

Unified memory lets capable models operate locally. As memory becomes local, intelligence becomes more persistent, private and personal.

200B

Frontier intelligence, human scale / 前沿智能进入人类尺度

Models once associated with large infrastructure can increasingly be tested or operated within compact systems.

130+ / 14+

Intelligence becomes modular / 智能正在模块化

Different tasks, latency, privacy and economics require different models. Advantage shifts to orchestrating the right intelligence at the right moment.

APP → AGENT

Navigation gives way to intention

Applications wait to be opened and contain functions. Agents remain oriented toward objectives and coordinate capabilities.

UI → JIT

Temporary interface, continuous context

Agent-native systems generate only the information and controls required by the present moment. 界面变得临时,情境变得连续。

1 → ∞

From device to constellation / 设备星群

A coordinated field of local processors, sensors, surfaces, agents and services composes itself around the person.

MANY

Intelligent pluralism / 智能多元化

The future is not one device for everyone, but many forms of intelligence for many ways of living.

04 / Five-year horizon

2026—2031: hypotheses about the direction of the system.

These are not predictions of individual products. Countercurrent studies the structural forces that may define the next computing cycle.

这些不是对具体产品的预测,而是关于整个系统前进方向的假设。

2026—2027

Local Intelligence Becomes Practical / 本地智能开始具备实用性

Personal AI computers evolve into local intelligence infrastructure for privacy, persistence, high-frequency inference and continuous operation. Local compute becomes the private ground beneath cloud intelligence.

2027—2028

The Application Layer Begins to Dissolve / 应用层开始溶解

Users express goals without choosing the app, menu or workflow. Applications become capabilities invoked by agents; operating systems govern agents, permissions, memory, identity and execution.

2028—2029

Interfaces Become Temporal / 界面变得具有时间性

Voice, vision, touch, movement, presence and environmental information form one interaction field. The best interface appears at precisely the right moment.

2029—2030

Intelligence Becomes Portable Across Objects / 智能开始跨物件迁移

Context, preferences and long-term memory become less dependent on one device. Identity, memory, permissions and agent portability become strategic terrain.

2030—2031

Computing Fragments Into Meaningful Forms / 计算分化为更有意义的形态

New objects emerge for professions, communities, rituals, environments and expression—creating a new material culture of intelligence chosen for meaning as well as capability.

05 / Principles

The future should be plural.

Intelligence should approach the human. / 智能应当靠近人。
Technology should reduce interface, not multiply it. / 技术应当减少界面。
Autonomy without agency is only automation. / 不赋予人选择权的自主性,只是自动化。
Privacy is an architecture, not a setting. / 隐私是一种架构。
A device should belong to a life, not a demographic. / 设备应当属于一种具体生活。
The future should be plural. / 未来不应只有一种标准答案。
06 / Strategic conviction

The next great computing company may begin by defining a new relationship.

Between human intention and machine action; local privacy and distributed intelligence; autonomy and human authority; universal capability and individual expression.

人的意图与机器的行动;本地隐私与分布式智能;自主执行与人类权力;通用能力与个体表达。

Countercurrent exists to investigate this relationship before it becomes obvious.

Countercurrent 的存在,是为了在这一关系成为共识之前,先行研究它。