分身如何成长How the twin grows 技术内核Twin core 隐私与掌控Privacy & control 咨询与成长Consultation & learning 常见问题Q&A 创建专家分身Build your expert twin
个人 AI 分身与隐私基础设施PERSONAL AI TWIN & PRIVACY INFRASTRUCTURE

你积累的经验判断，
从此可以分身。 Your experience,
now with a twin.

用工作日志、项目和真实案例训练你的 AI 分身。当 AI 生成的代码或方案需要专业判断时，让分身带着上下文向真人专家咨询：你解决问题，分身获得经授权的专家经验；专家获得收入，也从真实咨询中完善自己的分身。最终，让经验丰富的分身与知识丰富的 AI 协作，把时间还给创造与生活。 Train your AI twin on work logs, projects and real cases. When AI-generated work needs expert judgment, your twin can consult a human expert with full context. You solve the problem while your twin gains authorized expertise; the expert earns income and improves their own twin. Over time, experience-rich twins work with knowledge-rich AI—giving you more time to create and live.

发起咨询Start a consultation→AI 结果拿不准AI result uncertain
申请成为专家Apply as an expert有专业经验Professional experience
上传资料，训练分身Upload material and train your twin →

一个用户 · 一个长期分身 · 每次成长都由你确认One user · one persistent twin · every improvement approved by you

下滑SCROLL
一个用户 · 一个分身 — 个人资产 — 有界运行 — 三域隔离 — 专家确认 — 评测发布 — 版本回退 — 一个用户 · 一个分身 — 个人资产 — 有界运行 — 三域隔离 — 专家确认 — 评测发布 — 版本回退 — ONE USER · ONE TWIN — PERSONAL ASSETS — BOUNDED RUNTIME — THREE-DOMAIN ISOLATION — EXPERT CONFIRMATION — EVALUATED RELEASES — ROLLBACK — ONE USER · ONE TWIN — PERSONAL ASSETS — BOUNDED RUNTIME — THREE-DOMAIN ISOLATION — EXPERT CONFIRMATION — EVALUATED RELEASES — ROLLBACK —
付费咨询 · 双方成长PAID CONSULTATION · BOTH SIDES GROW

真正让分身成长的，
是一个个真实问题。What truly trains a twin
is a stream of real questions.

如果你是某个行业的专家，每一次咨询都会带来新的问题、案例、边界和结果。你对分身初稿的修改、判断依据和最终交付，是训练专家分身最有价值的材料。If you are an expert in a field, every consultation brings new questions, cases, boundaries and outcomes. Your edits to the twin’s first draft, the reasoning behind your judgment and the final delivery are the most valuable training material for an expert twin.

发起咨询的人也不只是得到一次答案。专家明确交付的方法、判断标准和检查步骤，经授权后可以进入咨询者自己的分身，成为以后能够再次使用的经验。The person seeking advice receives more than a one-off answer. With authorization, methods, judgment criteria and review steps explicitly delivered by the expert can enter the requester's twin as reusable experience.

提供咨询的人CONSULTING EXPERT01

获得收入，也把经验留在分身里Earn income—and retain the experience in your twin

你的分身先整理材料、分析问题并生成初稿；你负责关键判断、修改和最终交付。Your twin first organizes the material, analyzes the question and drafts a response; you make the consequential judgments, edit the draft and approve the final delivery.

分身整理与初稿TWIN PREPARES→专家判断与修改EXPERT JUDGES→最终交付FINAL DELIVERY

咨询结束后，系统会从你的修改中整理出：After the consultation, the system extracts from your edits:

  • 新的真实案例New real cases
  • 新发现的风险和边界Newly identified risks and boundaries
  • 分身遗漏或判断错误的地方What the twin missed or judged incorrectly
  • 可以重复使用的检查方法Reusable review methods
  • 用于检验分身的新问题New questions for testing the twin

经过你的确认和测试，这些经验会进入分身。咨询做得越多，分身越能理解你的判断，并逐步承担更多准备工作。After you confirm and test them, these candidate experiences enter your twin. The more consultations you complete, the better your twin understands your judgment and the more preparation work it can take on.

发起咨询的人CONSULTATION REQUESTER02

解决当前疑问，也让分身学会处理同类问题Solve today’s uncertainty—and teach your twin to handle similar questions

在使用 AI 的过程中，当你遇到拿不准的判断与选择，可以发起专家咨询。一次付费咨询，不只解决当前问题，也让你的分身有机会学会处理同类问题。When AI leaves you with an uncertain choice, start an expert consultation. A single paid consultation can solve the current problem and give your twin a chance to learn how to handle similar questions.

你得到的不只是一份专家意见，还可以获得专家明确交付或允许使用的：Beyond an expert opinion, you can receive methods explicitly delivered or authorized by the expert:

  • 分析问题的方法Methods for analyzing the problem
  • 判断时关注的关键条件Key conditions to examine
  • 可以重复执行的检查步骤Repeatable review steps
  • 方法适用和不适用的范围Where the method does and does not apply
  • 什么时候必须再次请专家确认When expert confirmation is still required

这些内容经过授权和测试后，可以进入你的分身。下一次遇到类似问题，它可以先完成分析和检查，不必再次从零开始。After authorization and testing, these can enter your twin. The next time a similar question appears, it can perform the initial analysis and checks instead of starting from zero.

ONE CONSULTATION / TWO RECORDS

一次咨询，产生两份不同的成长记录One consultation creates two distinct growth records

专家的分身EXPERT’S TWIN

获得收入Earns income

获得新的案例、修改和边界。Gains new cases, corrections and boundaries.

SEPARATE
咨询者的分身REQUESTER’S TWIN

解决当前问题Solves the current problem

获得经授权的方法、规则和经验。Gains authorized methods, rules and experience.

双方分别确认、分别测试、分别写入自己的分身。Each side confirms, tests and writes to its own twin separately.

01 / VISIBLE立即看见变化See the change immediately

咨询结束后，双方马上看到各自的分身学到了什么。After the consultation, both sides immediately see what their own twin learned.

02 / VERIFIED用新问题验证Validate with a new question

不是只保存资料，而是用相似任务验证分身能否正确使用。Knowledge is not merely stored; a similar task tests whether the twin can apply it correctly.

03 / OWNED经验各自归属Experience stays with its owner

客户原始资料不会自动进入专家分身，专家未授权的方法也不会自动转移。Raw client material does not automatically enter the expert’s twin, and an expert’s methods never transfer without authorization.

普通咨询解决一次问题；Twindoo 在解决问题的同时，让双方各自沉淀一项可复用的能力。Ordinary consultation solves one problem. Twindoo solves the problem and leaves each side with a reusable capability.

§ 01第一次训练 · 可验证的成长First training session · verifiable growth

一份资料，看见分身如何获得新能力。One source. See how your twin gains a new capability.

资料不会被静默存入知识库。Twindoo 会展示发现的候选知识、规则、Skill 与能力边界，用新旧结果对比验证变化，再由你决定是否写入正式分身。Your material is not silently stored in a knowledge base. Twindoo shows the candidate knowledge, rules, skills and boundaries it found, validates the change with a before-and-after test, and lets you decide what enters the live twin.

已授权材料AUTHORIZED SOURCE

《AI Agent 上线审查记录》AI Agent Launch Review

系统从一份已获授权的真实材料中提炼候选能力，并标注来源、适用任务与不适用范围。The system extracts candidate capabilities from one authorized real-world source and labels provenance, applicable tasks and exclusions.

候选 SkillCandidate skill1 项1 候选规则Candidate rules4 条4 能力边界Capability boundary1 条1 测试案例Test cases2 个2
BEFORE / 训练前BEFORE / PRE-TRAINING

只检查了权限控制Checks access control only

“方案具备权限控制，可以进入上线准备。”“The plan has access controls and can move into launch preparation.”

AFTER / 训练后AFTER / POST-TRAINING

开始执行完整上线审查Runs a complete launch review

“当前方案尚不适合上线。除权限控制外，还缺少状态持久化、断点恢复、回滚和完整审计记录。”“The plan is not ready to launch. Beyond access control, it still lacks state persistence, recovery, rollback and complete audit records.”

确认写入CONFIRM 修改后写入EDIT & SAVE 暂时保留HOLD 重新测试RETEST 拒绝学习REJECT
§ 02分身技术内核The twin core

一个用户，一个长期分身。
能力进化，掌控不变。One user, one persistent twin.
Capabilities evolve. Control stays with you.

同一个分身长期积累你的身份与偏好、判断规则、Skills、知识、真实经验、评测标准、运行策略和版本。不同任务和专业能力会成为它的不同 Skills，而不是创建多个彼此割裂的身份。更换模型，不会抹掉你已经建立的分身。The same twin accumulates your identity and preferences, judgment rules, skills, knowledge, real experience, evaluation criteria, operating policies and versions over time. Different tasks and professional capabilities become different skills within that same twin rather than creating multiple fragmented identities. Changing the model does not erase the twin you have built.

个人资产Personal assets档案 · 规则 · 能力 · 知识profile · rules · skills · knowledge
确认经验Confirmed experience真实任务 · 专家纠正 · 评测real work · corrections · evals
受控运行Controlled runtime授权 · 边界 · 工具 · 人工审批authorization · boundaries · tools · approval
版本系统Version system确认 · 发布 · 追踪 · 回退confirm · release · trace · rollback
Twindoo 专家分身技术内核：唯一入口、有界运行时、能力层、分层记忆，以及专家确认和评测后发布的成长路径 Twindoo expert AI twin core: single gateway, bounded runtime, capabilities, layered memory, and releases that require expert confirmation and evaluation
TWIN CORE / CONTROLLED BY DESIGN

所有请求先经过唯一入口校验身份、授权、期限与任务边界；运行时只读取当前被授权的上下文。任何一步缺失关键条件，系统都会拒绝继续执行。Every request first passes a single gateway that checks identity, authorization, deadline and task boundary. The runtime reads only currently authorized context. Missing a required condition stops execution.

训练的真实含义What training means

训练分身，不等于训练模型参数。Training your twin does not mean training model weights.

在当前阶段，训练更新的是你的个人资产、判断规则、经验与评测。系统先形成候选内容，只有经过你的确认、测试与发布，才会进入正式分身版本。At this stage, training updates your personal assets, judgment rules, experience and evaluations. The system proposes candidates; only your confirmation, testing and release can add them to the live twin.

V1 工作边界V1 work boundary

AI 先做，关键判断由本人确认。AI does the groundwork. You confirm consequential judgments.

分身可以整理材料、检索知识、执行已确认的 Skill、标记证据与风险，并生成分析和交付初稿；高风险、不可逆或超出边界的判断，仍由本人确认并正式交付。The twin can organize material, retrieve knowledge, run confirmed skills, flag evidence and risks, and draft analyses or deliverables. High-risk, irreversible or out-of-scope judgments still require your confirmation and formal delivery.

§ 03分身成长Twin growth

四步，让分身从认识你，到真正替你工作。Four steps—from getting to know you to doing real work for you

经过确认的新能力，都会以 Skills 的形式进入同一个长期分身，持续叠加。Once confirmed, new capabilities join the same persistent twin as skills, accumulating over time.

01 / MATERIAL本人资料Your material
02 / CORRECTION本人纠正Your corrections
03 / WORK真实任务Real work
04 / CONSULT发起或接受咨询Initiating or accepting consultations
05 / OUTCOME结果反馈Outcome feedback
expert twin / growth path
认识你的专业Learn your expertise
文档 · GitHub · 工作记录Docs · GitHub · work records
学习判断标准Learn your standards
专家纠正 · 评测Corrections · evaluations
AI
承担真实任务Take on real work
范围明确 · 稳定交付Clear scope · consistent delivery
服务LIVE
进入更多场景Enter more scenarios
Agent · 工作流 · 平台Agents · workflows · platforms
01 — 文档 · GitHub · 工作记录 · 案例 · 方法论DOCUMENTS · GITHUB · WORK RECORDS · CASES · METHODOLOGIES

第一步：让分身认识你的专业Step 1: Let the twin learn your expertise

导入文档、GitHub、工作记录、案例与方法论，建立属于你的个人能力底座。Import documents, GitHub repositories, work records, cases and methodologies to establish your personal capability foundation.

02 — 专家纠正 · 评测 · 判断边界EXPERT CORRECTIONS · EVALUATIONS · JUDGMENT BOUNDARIES

第二步：把判断标准教给分身Step 2: Teach the twin your judgment standards

通过专家纠正与评测，确认它理解哪些方法可以使用、哪些边界不能越过。Use expert corrections and evaluations to confirm which methods it can apply and which boundaries it must never cross.

03 — 明确场景 · 真实任务 · 稳定交付CLEAR SCENARIO · REAL TASKS · CONSISTENT DELIVERY

第三步：让分身开始承担真实任务Step 3: Put the twin to work on real tasks

从一个范围明确的专业场景开始，验证它能否稳定完成工作。Start with one clearly scoped professional scenario and verify that it can deliver consistently.

04 — API · Agent · 工作流 · 其他平台API · AGENTS · WORKFLOWS · OTHER PLATFORMS

第四步：把同一个分身带到更多场景Step 4: Bring the same twin into more scenarios

通过受控接口连接更多 Agent、工作流与其他平台。Connect it to more agents, workflows and platforms through controlled interfaces.

每天几分钟，完成一次有结果的训练。A few focused minutes. A training session with a visible result.

系统根据近期工作、咨询反馈和能力缺口，准备短时训练任务。每次训练结束后，你都会看到发生了哪些变化、通过了哪些测试，以及哪些任务会受到影响。Based on recent work, consultation feedback and capability gaps, the system prepares short training tasks. After each session, you can see what changed, which tests passed and which tasks are affected.

01 / CONFIRM确认、修改或拒绝一条候选规则Confirm, edit or reject a candidate rule
02 / COMPARE比较两个结果，选择更符合你的判断Compare two results and choose the one that matches your judgment
03 / BOUNDARY补充一个能力边界或转本人条件Add a capability boundary or escalation condition
04 / TEST用相似案例验证刚确认的规则Validate a confirmed rule on a similar case
§ 04隐私与掌控 · 核心Privacy & control · the core

隐私，是分身成立的前提。Privacy is what makes a twin possible.

你的个人资产、当前任务材料与待确认的经验，分别进入三个隔离数据域；未经授权，不会跨域流动。Your personal assets, current task material and unconfirmed experience live in three isolated data domains; without authorization, nothing moves across them.

DOMAIN / 01

个人分身域Personal twin domain

存放你已经确认并发布的规则、技能、知识、案例、评测和版本。只属于你的分身。Stores your confirmed and released rules, skills, knowledge, cases, evaluations and versions. It belongs only to your twin.

DOMAIN / 02

任务工作区Task workspace

存放当前订单或任务经授权的材料与状态。任务结束后不会自动沉淀为个人资产。Holds material and state authorized for the current order or task. It does not automatically become a personal asset when work ends.

DOMAIN / 03

候选经验域Candidate experience domain

存放系统从真实工作中提议的经验、纠正与规则。未确认、未评测，就不能进入正式分身。Holds proposed lessons, corrections and rules from real work. Nothing enters the live twin before confirmation and evaluation.

训练必须授权Training requires authorization

任何外部材料在进入正式分身前，都必须经过以下流程：Before any external material can enter the live twin, it must pass through:

授权AUTHORIZE→解析PARSE→脱敏DE-IDENTIFY→质检VALIDATE→候选资产CANDIDATE→专家确认CONFIRM→评测EVALUATE→发布版本RELEASE
01

服务边界由你定义You define the boundary

你决定分身能处理什么、何时拒绝，以及何时必须交由本人处理。You decide what the twin can handle, when it must refuse and when the work must return to you.

02

可追踪、可回退、可退出Traceable, reversible and exit-ready

每项资产保留来源、用途与版本；你可以审阅变化、回退版本，并按协议导出允许的资产包或完成退出删除。Every asset retains its provenance, purpose and version. You can review changes, roll back and, under the agreement, export the permitted asset package or complete the exit and deletion process.

03

私人资料不默认用于训练公共模型Private material is not used to train public models by default

私人资料和咨询原文不会默认用于训练公共模型、平台公共能力或其他用户的分身。Private material and raw consultation content are not used by default to train public models, shared platform capabilities or another user's twin.

§ 05从现在到未来From now to what comes next

同一个分身，逐步承担更长期的工作。The same twin takes on longer-horizon work over time.

咨询是 Twindoo 首批落地的场景之一，不是分身的最终形态。随着能力经过更多真实任务验证，同一个分身会在始终受控的前提下进入更多工作流。Consultation is one of Twindoo's first live applications—not the twin's final form. As capabilities are validated across more real tasks, the same twin can enter more workflows while remaining controlled.

NOW / 当前NOW

协助与受控交付Assistance with controlled delivery

整理、检索、分析和生成初稿；关键判断与正式交付仍由本人确认。Organize, retrieve, analyze and draft, while consequential judgments and formal deliverables still require your confirmation.

NEXT / 下一步NEXT

执行跨日任务Run multi-day tasks

在明确目标、权限与预算下保存状态和检查点，失败后恢复，并在异常时通知本人。Work within explicit goals, permissions and budgets, preserving state and checkpoints, recovering from failure and alerting the owner when needed.

LONG TERM / 长期LONG TERM

连接更多 Agent 与平台Connect to more agents and platforms

让经过验证的低风险能力进入更多工作流，并把新经验带回同一个长期分身。Bring validated low-risk capabilities into more workflows and return new experience to the same persistent twin.

§ 06专业咨询 · 服务与履约Expert consultation · service and fulfillment

你专注专业判断，平台处理服务流程。You focus on expert judgment. Twindoo handles the service flow.

你定义服务范围、价格和需要本人参与的边界；Twindoo 负责整理需求、匹配服务，以及交易与履约。You define the service scope, price and when your involvement is required. Twindoo clarifies the request, matches the right service, and handles transactions and fulfillment.

01

需求先被整理清楚Demand is clarified first

平台先确认问题目标、材料完整度与服务适配度，再交给合适的分身或专家。Twindoo confirms the goal, material readiness and service fit before routing it to the right twin or expert.

已整理READY
问题与材料task context
02

服务规则由你定义You define the service rules

你决定可以提供哪些服务、如何定价，以及哪些情况需要本人参与。You decide which services to offer, how to price them and when you must be involved.

由你定义EXPERT-SET
服务与边界service & boundary
03

交易与履约由平台处理Twindoo handles transactions and fulfillment

Twindoo 负责合同、收款、服务切换与争议流程，让专家专注于专业交付。Twindoo manages contracts, payments, service handoffs and dispute workflows so experts can focus on professional delivery.

平台处理HANDLED
交易与履约transactions & fulfillment
§ 07Q&A

专家分身常见问题Expert AI twin Q&A

专家分身和普通 AI 助手、数字人有什么不同？How is an expert AI twin different from a generic AI assistant or digital human?

普通 AI 助手使用通用模型生成回答，数字人偏形象呈现。Twindoo 专家分身由一位专家的规则、案例、评测与边界持续定义，是可确认、可追溯、可迭代的个人专业系统。A generic AI assistant generates responses using a general-purpose model, while a digital human mainly presents a persona. A Twindoo expert AI twin is continuously defined by one expert's rules, cases, evaluations and boundaries—a personal professional system with expert confirmation, traceability and continuous iteration built in.

所谓“训练分身”，是在微调一个属于我的大模型吗？Does “training my twin” mean fine-tuning a model for me?

当前阶段不是。Twindoo 更新的是与你绑定的规则、知识、技能、经验、评测与运行策略，而不是静默修改模型参数。个人资产与基座模型相对解耦，因此分身可以持续积累，也能按任务选择合适的模型。Not at this stage. Twindoo updates the rules, knowledge, skills, experience, evaluations and operating policies bound to you—not model weights in the background. Personal assets remain relatively independent of the foundation model, so the twin can persist while using the right model for each task.

发起或接受咨询后，分身会学到什么？What does the twin learn after initiating or accepting a consultation?

双方各生成一份相互隔离的候选能力记录：专家侧来自真实问题、修改和边界；咨询者侧只包含专家明确交付或授权的方法。两份记录都必须分别确认、测试和发布，未授权内容不会转移。Each side receives a separate, isolated candidate capability record. The expert's record comes from real questions, edits and boundaries; the requester's contains only methods explicitly delivered or authorized by the expert. Each record must be independently confirmed, tested and released. Unauthorized content never transfers.

我的客户材料会被拿去训练吗？Will client materials be used for training?

默认不会。客户原始材料只服务当前任务；未经明确授权、脱敏与评测，不会进入你的正式分身，也不会默认用于训练公共模型或其他用户的分身。Not by default. Raw client materials serve only the current task. Without explicit authorization, de-identification and evaluation, they do not enter your live twin and are not used by default to train public models or another user's twin.

分身答错了，或遇到超出边界的任务怎么办？What happens when the twin is wrong or a task exceeds its boundary?

回答保留版本、证据引用与运行记录。高风险、不可逆或超出边界的任务，分身会拒答或带着上下文转交本人；V1 的关键判断与正式交付仍由真人确认。Responses retain version, evidence references and runtime records. The twin refuses high-risk, irreversible or out-of-scope work, or escalates it back to you with the full context attached. In V1, consequential judgments and formal deliverables still require human confirmation.

平台会不会复制我、压低我的价格？Will Twindoo clone me or undercut my price?

Twindoo 不会使用你的资产训练与你直接竞争的分身；分身服务范围与本人服务价格由你参与定义，并在专家协议中明确。Twindoo does not use your assets to train a twin that directly competes with you. Your twin's service scope and the pricing of your direct services are defined with you and documented in the expert agreement.

用一份真实材料，建立分身的第一项能力。Build your twin's first capability from real material.

从一个边界明确的真实任务开始。Twindoo 会与你一起提炼候选知识、规则和 Skill，完成第一次测试，并建立可查看、可评测、可回退的分身版本。Start with one clearly bounded real task. Twindoo helps extract candidate knowledge, rules and skills, run the first test, and establish a twin version that is inspectable, evaluable and reversible.

在明确领域拥有真实项目经验Real project experience in a defined field
拥有可验证的成果、作品或案例Verifiable outcomes, work or cases
愿意用代表性材料训练分身Willing to train with representative material
愿意定义边界并验证结果Willing to define boundaries and validate results
创建专家分身Build your expert twin

适合在明确方向上拥有真实项目经验与可验证成果的专家For experts with real project experience and verifiable outcomes in a defined field

功能即将上线Coming soon我们正在准备首批体验，敬请期待。We are preparing the first release. Stay tuned.