面向专利专业人员，把起草、分析、审查答复、证据研究与知识管理等任务连接在一个在线工作空间中。Built for patent professionals, it connects drafting, analysis, prosecution response, evidence research, and knowledge management in one online workspace.
从发明交底、权利要求与标准分析，到审查意见答复与团队知识沉淀，工作不再分散在不同工具中。From invention disclosure, claims, and standards analysis to Office Action response and team knowledge capture, work no longer has to be scattered across separate tools.
后台理解技术与法律内容的含义，帮助提取重点、连接证据、比较差异并形成可用洞察。The AI-powered backend understands technical and legal meaning to extract key points, connect evidence, compare differences, and produce usable insights.
前端针对每项任务突出最重要的证据、争点与下一步，让专利专业人员更快看懂并使用数据。For each task, the frontend highlights the most important evidence, issues, and next actions—helping patent professionals understand and use data faster.
理解复杂技术材料，提炼核心发明点，发现信息缺口，并让发明人与专利团队围绕同一创新主线协作。Understand complex technical material, surface the inventive contribution, identify information gaps, and align inventors with the patent team.
围绕商业重点组织发明概念、独立与从属权项、说明书支持和法域要求，并通过专家审阅完成申请文件。Organize inventive concepts, independent and dependent claims, specification support, and jurisdiction requirements around business priorities, with expert review before completion.
将每个权项要素与技术标准、目标产品或其他证据逐项映射，并把重要结论连接到可复核的原始资料。Map each claim element against technical standards, target products, or other evidence, with important conclusions tied to reviewable source material.
汇总驳回理由、审查历史、现有技术与法律依据，比较多种答复策略，并生成经过复核与验证的完整答复文件。Bring together rejections, prosecution history, prior art, and legal authorities; compare response strategies; and produce a reviewed, validated response document.
在同一视图中比较同族申请的权项、法域差异、审查结果与保护范围，快速发现一致性问题和组合机会。Compare claims, jurisdiction differences, prosecution outcomes, and protection across a patent family to reveal consistency issues and portfolio opportunities.
把技术材料提炼成清晰的创新点与商业价值。Turn technical material into clear inventive concepts and business value.
围绕核心贡献设计更有层次的权利要求。Shape layered claims around the invention’s core contribution.
快速看清权项与技术标准之间的证据关系。Reveal the evidence relationship between claims and technical standards.
把驳回、现有技术与答复策略放在同一视图中。Bring rejections, prior art, and response strategy into one coherent view.
比较专利族差异，快速理解授权结果。Compare patent-family differences and understand grant outcomes quickly.
让每次专家定稿都反哺下一次工作。Use every expert-approved result to improve the next matter.
优势不只来自模型能力，更来自围绕专业质量设计的工作流架构、界面与统一平台。The advantage comes not only from the model, but from workflow architecture, interface design, and a unified platform built around professional quality.
真实资料接地、多角色审阅、反馈循环与交付前验证，共同降低错误进入最终成果的风险。Source grounding, role-separated review, feedback loops, and pre-delivery validation reduce the risk of errors reaching the final output.
多路径回退、自愈与阶段恢复，让单个工具、来源或步骤的失败不会拖垮整个任务。Fallback routes, self-healing, and stage recovery prevent one failed tool, source, or step from collapsing the whole task.
全员在统一平台工作，把标准、输出、审计与用户反馈汇成可持续提升质量的数据基础。A shared platform turns standards, outputs, audits, and user feedback into a foundation for continuous quality improvement.
任务型界面把案件信息、证据、策略选项与下一步集中呈现；引导式流程让用户轻松完成审查意见答复起草。The task-focused UI puts matter information, evidence, strategy options, and next steps in one view, while a guided workflow makes Office Action response drafting easy to navigate.
真实案件示例：系统把分散在多轮审查文件中的时间线、权项状态、驳回依据、争议焦点与下一步机会汇总到同一页面。Real matter example: the platform brings the timeline, claim status, rejection grounds, disputed issues, and next opportunities from multiple prosecution rounds into one page.
Magri 被用于 §102 驳回；申请人随后修改权项 1、10、12、20，强调 DSP / 纠错电路数据。Magri supported the §102 rejection; the applicant then amended claims 1, 10, 12, and 20 to emphasize DSP / error-correction data.
审查员维持对 “digital signal processor” 的宽泛解释，并对权项 20 新增书面描述问题。The examiner maintained a broad reading of “digital signal processor” and raised a new written-description issue for claim 20.
全部 20 项权利要求仍待处理，当前需要选择继续审查路径。All 20 claims remain pending; a path for continued prosecution must now be selected.
专业人员控制关键输入与策略选择；AI 在后台完成资料汇总、分析、起草、复核、验证与文档输出。Professionals control the critical inputs and strategic choice; AI handles collection, analysis, drafting, review, validation, and document production.
Claims 1, 12, and 20 are currently amended.
§102(a)(1): Magri’s monitoring state machine does not perform the claimed data-extraction processing.
Reconsideration and allowance of the pending claims are respectfully requested.
简单 ChatGPT 或单一 Agent 循环往往把生成与判断交给同一个模型；平台把检索、分析、起草、审阅与验证拆成相互制衡的阶段。A simple ChatGPT session or single-agent loop often leaves generation and judgment to the same model. The platform separates retrieval, analysis, drafting, review, and validation into checks and balances.
关键结论从法律资料、标准与案件证据中检索，而不是依赖模型记忆。Key conclusions are retrieved from legal authorities, standards, and matter evidence—not model memory.
提议者与评审者分离，通过质疑、比较与反馈循环修正薄弱方案。Proposers and reviewers are separated so critique, comparison, and feedback can correct weak approaches.
引用、法律依据、权项一致性与形式要求在成稿前接受结构化检查。Citations, legal support, claim consistency, and formal requirements receive structured checks before delivery.
简单循环中，一个链接失效、文件抓错或工具报错就可能迫使任务重来；平台把故障检测、替代检索、验证与局部重跑设计进流程。In a simple loop, one dead link, wrong document, or tool failure can force a restart. The platform builds detection, alternate retrieval, verification, and targeted reruns into the workflow.
同一资料可通过多个权威与公开来源获取，减少单一接口依赖。The same material can be retrieved through multiple authoritative and public sources, reducing single-interface dependency.
发现缺失或错误引证后，自动寻找替代、验证并让对应分析重新执行。When a citation is missing or wrong, the platform finds and verifies a replacement, then reruns the relevant analysis.
任务与单次连接解耦；中断后从已有结果继续，无需整条流水线从头开始。Work is decoupled from one connection and can resume from completed results instead of restarting the entire pipeline.
集中化不只是使用方便；它让组织拥有全部输出、运行轨迹与用户反馈，从而把质量保证从抽样检查升级为持续运行。Centralization is more than convenience. Access to outputs, run histories, and user feedback turns quality assurance from occasional sampling into a continuous operation.
统一规则、模板、审阅步骤与法域要求，减少人员之间的质量波动。Shared rules, templates, review stages, and jurisdiction requirements reduce quality variation between employees.
AI 可扫描答复、输出与运行记录，自动发现遗漏、矛盾、引用或合规问题。AI can scan responses, outputs, and run records to detect omissions, inconsistencies, citation issues, and compliance risks.
跨案件聚合问题模式，定位经常返工的环节与系统性风险。Patterns across matters reveal recurring rework, workflow pain points, and systemic risks.
把高频质量问题转化为具体培训主题、示例与辅导重点。Turn recurring quality issues into concrete training topics, examples, and coaching priorities.
将外部代理起草的审查意见答复送入同一流水线，识别质量差异与潜在问题。Run outside-counsel Office Action drafts through the same review pipeline to identify quality gaps and potential issues.
分析聊天记录、修改与反馈，把真实用户需求提炼为规则、流程与产品改进。Analyze chat transcripts, edits, and feedback to distill real user needs into better rules, workflows, and product capabilities.
平台把专家的修改与反馈转化为可读经验；经人确认后，再帮助下一次工作。The platform turns expert edits and feedback into readable lessons. After human approval, those lessons help the next matter.
AI 把重要差异整理成一条简单、可读的候选经验。AI turns an important difference into one simple, readable candidate lesson.
审阅者可批准、编辑或拒绝。未经确认，任何经验都不会自动生效。A reviewer approves, edits, or rejects the lesson. Nothing changes automatically without confirmation.
经确认的经验会帮助团队处理后续案件。Approved lessons help the team handle future matters.
每个案件先使用同一质量基线，再叠加对应法域的必需规则。以下仅展示简化节选。Every matter starts with the same quality baseline, then adds the required local rules. Simplified excerpts are shown below.
规则节选Rule excerpt“核对原文，不编造；完整回应争点；表达清晰、专业。”“Verify sources; never invent. Address every issue. Write clearly and professionally.”
规则节选Rule excerpt“权项修改使用规定标记；答复符合 USPTO 程序与文件惯例。”“Use required claim markings; follow USPTO procedure and document conventions.”
规则节选Rule excerpt“创造性使用问题—解决方法；修改需要直接、明确的依据。”“Use the problem-solution approach; amendments require direct and unambiguous support.”
规则节选Rule excerpt“使用 CNIPA 本地术语；按中国修改规则与答复格式处理。”“Use CNIPA-native terminology; follow China-specific amendment and response rules.”
规则节选Rule excerpt“使用国际阶段要求；围绕新颖性、创造性与单一性组织答复。”“Apply international-stage requirements; organize responses around novelty, inventive step, and unity.”
| 关键维度Dimension | ChatGPT / 简单 Agent 循环ChatGPT / simple agent loop | 编排式集中平台Orchestrated centralized platform |
|---|---|---|
| 正确性Correctness | 同一模型生成并自检，常依赖模型记忆One model generates and self-checks, often relying on model memory | 真实来源接地 + 独立审阅 + 反馈循环 + 验证门禁Source grounding + independent review + feedback loops + validation gates |
| 稳健性Robustness | 一个工具或链接失败，任务停止或重来One failed tool or link can stop or restart the task | 多路径回退 + 自愈 + 局部重跑 + 断点恢复Fallback routes + healing + targeted reruns + recovery |
| 集中化Centralization | 对话与输出分散在个人账户，组织不可见Chats and outputs remain scattered in personal accounts | 全员工作汇入统一规则、权限、审计与数据体系Employee work shares one system for rules, access, audit, and data |
| 质量提升Quality improvement | 问题与用户反馈难以跨案件汇总Problems and feedback are difficult to aggregate across matters | 自动审计、模式识别、培训洞察与系统改进形成闭环Automated audit, pattern detection, training insights, and system improvement form a closed loop |
以真实资料、独立审阅、反馈循环与验证降低错误风险。Reduce error risk through real sources, independent review, feedback, and validation.
用回退、自愈与恢复机制，让复杂任务在现实故障中继续运行。Use fallbacks, healing, and recovery to keep complex work moving through real-world failures.
把全组织的输出与反馈转化为一致标准、自动质检与持续改进。Turn organization-wide outputs and feedback into consistent standards, automated QA, and continuous improvement.