带来能力与需求Bring capabilities and needs
参与者说明设备、服务或智能体能做什么，以及希望改善的任务。Participants describe what their devices, services or agents can do and the task they want to improve.
把各自的能力，接到同一项真实任务。Bring different capabilities into one real task.
开发者提供软件能力，设备伙伴连接现场，场景方带来需求与反馈。AIOS 围绕任务组织调用，记录结果与贡献，让参与者有依据改进能力、讨论下一次协作。Developers provide software capabilities, device partners connect the field, and scenario owners bring needs and feedback. AIOS organizes calls around a task and records outcomes and contributions, giving participants a basis for improvement and future collaboration.
当前聚焦具体项目中的可信协作；更完整的能力网络与结算机制仍需运行证据支撑。The current focus is trusted collaboration in specific projects. Fuller capability networks and settlement mechanisms still need operating evidence.
围绕真实任务建立协作Collaboration around a real task
职责与机制示意Roles and mechanics
协作中心Shared focus
一项真实任务A real task
明确目标、边界与结果Clear goals, boundaries and outcomes
开发者Developers
提供可验证的能力Provide verifiable capabilities
智能体Agents
参与任务与能力调用Join tasks and capability calls
设备Devices
连接物理能力Connect physical capabilities
场景方Scenario owners
给出需求与现场反馈Bring needs and field feedback
合作伙伴Partners
提供资源与专业支持Provide resources and expertise
- 记录结果与贡献Record outcomes and contributions
- 吸收反馈与纠正Collect feedback and corrections
- 在下一项任务再验证Validate again in the next task
目标是让反馈可复用；改进效果仍需真实任务验证。The goal is reusable feedback; improvement still needs validation in real tasks.
从参与任务，到验证下一次改进From joining a task to validating the next improvement
在任务中调用Call capabilities in a task
按真实条件找到合适能力，在有效权限内组织执行。Find suitable capabilities under real conditions and organize execution within valid permissions.
记录实际结果Record actual outcomes
保留执行结果、失败原因、贡献来源与人的纠正，供后续核查。Keep outcomes, failure reasons, contribution provenance and human corrections for later review.
验证反馈是否有效Test whether feedback helps
用反馈评估能力版本，检查改进是否减少同类错误、是否可以再次使用。Use feedback to evaluate capability versions and check whether improvements reduce repeated errors and support reuse.
从反馈，回到下一次任务。From feedback to the next task.
一项任务，需要不同的参与方式。One task needs different kinds of participation.
人和组织提供能力、现场条件与需求；智能体和设备在任务中承担具体动作。先明确各自贡献与调用边界，再进入协作。People and organizations bring capabilities, field conditions and needs. Agents and devices perform specific actions in a task. Define contributions and calling boundaries before collaborating.
开发者提供软件能力Developers provide software capabilities
构建智能体、MCP 服务与 Skills，说明能力范围，并带入真实任务验证。Build agents, MCP services and Skills, define their scope and validate them in real tasks.
智能体参与任务执行Agents take part in execution
携带身份、状态与经验，按任务分工调用能力，并留下执行结果。Carry identity, state and experience, call capabilities according to task roles and leave execution outcomes.
设备伙伴提供物理能力Device partners provide physical capabilities
说明设备身份、位置、已有能力与调用条件，连接可信的现场动作。Define device identity, location, existing capabilities and calling conditions to support trusted field actions.
场景方提供需求与反馈Scenario owners provide needs and feedback
明确要改善的任务、空间条件、允许的动作，以及判断结果的方式。Define the task to improve, spatial conditions, permitted actions and how to assess outcomes.
伙伴提供专业与基础支持Partners provide specialist and infrastructure support
围绕具体项目提供渠道、制造、云模型或专业能力，确认各自参与范围。Provide channels, manufacturing, cloud and model resources or specialist capabilities for a specific project, with agreed participation scope.
先验证能否重复，再判断怎样扩展。Test repeatability before deciding how to expand.
当前检验执行可靠、失败恢复、持续使用与跨空间复用。城市级能力网络是长期方向；运营商可作为渠道与网络基础设施伙伴参与建设。Current validation examines reliable execution, failure recovery, sustained use and reuse across spaces. City-scale capability networks are a long-term direction; telecom operators may participate as channel and network-infrastructure partners.
检查执行与恢复Check execution and recovery
验证任务能否稳定完成，失败能否定位并恢复到安全状态。Test whether tasks can finish reliably and whether failures can be located and recovered to a safe state.
检查反馈与改进Check feedback and improvement
验证人的纠正是否进入版本评估，以及改进后同类错误是否减少。Test whether human corrections inform version evaluation and whether improvements reduce repeated errors.
检查跨空间复用Check reuse across spaces
验证状态模型、设备适配、权限策略和执行规则能否在第二个真实空间使用。Test whether state models, device adapters, permission policies and execution rules can be used in a second physical space.
说明你能做什么，找到一起验证的任务。Describe what you can do and find a task to test together.
明确设备、软件服务或智能体的能力、接入要求与调用边界，从具体协作中确认价值。Define your device, service or agent capability, integration requirements and calling boundaries, then test its value through concrete collaboration.