PERSONAL FLOW 拖动画布探索 · 点击节点阅读Drag to explore · Click to read
关于我About medonghuang
东煌 / Donghuang

Hi,我是东煌Hi, I’m Donghuang

我是个 AI BuilderI’m an AI Builder

深圳 · 中国Shenzhen · China

INFJ

喜欢找个安静的地方读书、骑车听歌,也爱看电影、随手拍下好风景。I like quiet places to read, cycling with music, a good film, and catching a beautiful view on camera.

LangChainAI AgentRAGMCPAgent SkillsPrompt EngineeringLoRACodexAI 产品设计AI product design工作流编排Workflow design用户研究User research产品原型Prototyping生成式内容Generative content创作者增长Creator growth
{
  "name": "Donghuang",
  "role": "AI Builder",
  "focus": "Creators × AI",
  "building": "Creator tools"
}
连接Connectchannels

保持连接。Stay connected.

想法、作品,以及下一次对话。Ideas, work, and our next conversation.

内心驱动Inner CallingInner Calling

探索 AI 时代,
个体如何创造,
又如何立足社会。
Exploring how individuals
create—and find their place—
in the age of AI.

我想参与构建一种以人为主体的 AI 协作方式:人提出问题、作出判断,与 AI 一起把想法变成作品和服务,再向真实世界交付价值。Creator,是我理解这场变化的起点。I want to help build human-led collaboration with AI: people frame problems and make judgments, work with AI to create products and services, and deliver value to the world. Creator is the identity through which I explore this change.

项目Projectswork

创作者与 AICreators × AI

思考ThoughtsKnow-how

从复杂问题出发,
让智能持续复利。
Start with complex problems.
Let intelligence compound.

下一站Next chapternext.chapter

从创作者开始,
探索新的工作形态。
Start with creators.
Explore new ways of working.

让一个人的知识、表达、判断与真实反馈,进入长期的人机协作。围绕具体创作与业务,探索人怎样提出问题,怎样与 AI 完成工作,又怎样把价值交付给别人。Bring a person’s knowledge, expression, judgment and real feedback into a lasting collaboration with AI. Explore how people frame problems, work with AI and deliver value through actual creative and business work.


也在研究这些问题?Working on these questions too?

如果你也在探索个体、AI 与组织之间的新关系,或正在具体的创作与业务中实践,欢迎交换经验,一起做些东西。If you are exploring new relationships between individuals, AI and organizations, or testing them in real work, I would like to exchange experiences and build together.

聊聊你正在探索的事Tell me what you’re exploring ↗
100%
DH / AI 时代,
人如何工作与创造?
In the age of AI,
how do people work and create?

关于人、AI 与创造On people, AI and creation

AI 时代,
人如何工作与创造?
In the age of AI,
how do people work and create?

从招聘到创作者产品,我一直在追问:一个人如何被需要,又如何主动创造自己的位置。From recruiting to creator products, I have kept asking how a person can be needed—and actively shape their own place.

我最早关心的,是人如何在社会中找到位置。It began with how people find a place in society.

在智联招聘时,我把“帮助人找到工作”看作一件值得长期做的事。一份工作连接着收入,也连接着一个人被需要的感受。后来,我接触过校招、社招、猎头协作、Web3 招聘和面向海外蓝领的 Agent。做这些产品时,我都在尝试把人与机会接得更近。At Zhaopin, helping people find work felt worth committing to. A job connects income with the feeling of being needed. I later worked across campus and experienced hiring, headhunter collaboration, Web3 recruiting and agents for overseas workers, trying to bring people closer to opportunity.

实践也让我看见了匹配的边界:平台可以改善信息和沟通,却不能独自解决岗位供给、技能结构与信任。换一种技术,并不意味着人就自然拥有了更多选择。我开始追问,除了进入一个已经存在的岗位,个体还能怎样参与社会、创造价值?Practice showed me the limits of matching. Better information and communication cannot alone resolve job supply, skills and trust. A new technology does not automatically create more choices. I began asking how people could participate in society and create value beyond filling an existing position.

如果工作可以交给 AI,人的位置该由什么定义?If AI can do the work, what defines a person’s place?

这并不是最近才出现的问题。2023 年,我已经在笔记里思考 AI 会怎样改变组织协作,也在整理“一人公司”背后的产品、运营、销售和交付。到 2024 年,我的问题变得更尖锐:如果越来越多任务不再依赖人的执行,原来以岗位和劳动交换收入的关系,会发生什么变化?This is not a new question for me. In 2023, my notes explored AI and organizational collaboration, alongside the product, operations, sales and delivery behind a one-person business. In 2024, I asked more directly what happens to the link between jobs, labor and income as more tasks can be performed without human execution.

早期判断里有焦虑,也有需要不断修正的推测。但那个问题一直留了下来:当完成任务的方式变了,人怎样继续被需要,又怎样主动决定自己要做什么?这是我希望通过产品实践持续探索的事。Those early reflections contained anxiety and predictions that need revisiting. The question remained: when task execution changes, how do people remain needed while choosing what they want to do? I want to explore it through building and real use.

Creator,是我对未来个体身份的一种理解。Creator is how I think about the future individual.

我现在所说的 Creator,范围比内容博主更广。它可以是写作者、设计师、开发者,也可以是发现一个问题、组织资源并做出解决方案的人。共同点在于:从自己的观察和意图出发,把某种东西创造出来,并让它对别人有用。Creator means more to me than a content publisher. It can be a writer, designer, developer, or someone who identifies a problem and organizes a solution. What connects them is turning observation and intention into something useful to others.

我倾向于把这种能力看作未来个体越来越重要的身份基础:能够提出值得做的事,借助 AI 组织创造,再把结果带回人与人的关系中。是否独立创业、是否在公司里,并不决定一个人能不能成为这样的 Creator。I see this capacity becoming increasingly important to individual identity: proposing worthwhile work, organizing creation with AI, and bringing the result back into human relationships. It can exist inside a company as well as in an independent practice.

人要留在回路里,而且要留在关键的位置。People need to remain in the loop, where it matters.

我理解的 Human in the loop,贯穿输入、协作与输出。人把真实世界的问题、经验、偏好和边界带进来;AI 参与检索、推演、生成与执行;人持续修正方向,对最终交付作出判断,并承担它进入社会后的责任。Human in the loop spans input, collaboration and output. People bring real problems, experience, preferences and boundaries. AI participates in retrieval, reasoning, generation and execution. People keep adjusting the direction, judging the deliverable and taking responsibility for its consequences.

因此,我更关心一种长期协作关系:AI 能否逐渐理解这个人,记住真实工作中的反馈,把一次次零散使用连接起来。人的经验进入系统,系统的产出又回到人的判断中,这种往返才有可能形成持续积累。That leads me toward a lasting collaborative relationship. Can AI learn about a person, retain feedback from real work and connect separate interactions? Human experience enters the system, while its output returns to human judgment. That repeated exchange is where cumulative progress may develop.

我过去做的产品,是这些问题的具体入口。The products I worked on gave these questions a concrete form.

FaceLive 让我探索一个人的表达、形象和服务如何通过 AI 延伸;在 Open Bot,我接触到知识、Agent 与业务渠道之间的连接。到了起点云和来画,我沿着短剧解说、漫影、红薯助手、带货与视频生成继续往前,关注一个想法怎样变成可交付的内容。FaceLive explored extending a person’s expression, image and service through AI. Open Bot connected knowledge, agents and business channels. At Qidianyun and Laihua, my work on commentary, animation, Xiaohongshu, commerce content and video generation examined how ideas become deliverable content.

MyyShop 又把问题推到了创作者与商家的关系里:作品做出来之后,别人是否需要它,合作是否产生价值,信任能不能延续?这些经历让我逐渐把人的表达、AI 协作和社会交付放到一起考虑。每个项目都只回答了其中一部分。MyyShop brought the question into creator-brand relationships: once something is made, does someone need it, does collaboration create value, and can trust continue? These experiences connected human expression, AI collaboration and delivery to others. Each project answered only part of the question.

新的工作方式,也需要新的协作关系。New ways of working need new forms of collaboration.

一个人拥有更多 AI 能力,并不会让客户、信任、责任和协作自动消失。我对“一人公司”的兴趣,也一直包含这些完整的业务职能。我想探索的是,个体能否围绕一个真实问题,调动 AI 和可信赖的伙伴,以更灵活的方式完成端到端的交付。Greater AI capability does not remove customers, trust, responsibility or collaboration. My interest in one-person businesses includes all those functions. I want to explore whether individuals can organize AI and trusted partners around real problems to deliver complete outcomes more flexibly.

这可能发生在独立业务里,也可能发生在小型团队或现有组织中。具体形式可以变化,但需要回到同一件事:谁在提出问题,谁在作判断,谁对最终价值负责。This could happen in an independent business, a small team or an existing organization. The form can vary, but the questions remain: who frames the problem, who makes judgments, and who is accountable for the value delivered?

我想探索的,是个体与 AI 共同创造的未来。I want to explore a future of individuals creating with AI.

我选择从创作者开始,是因为表达、受众、产品、收入和反馈在这里有具体的样子。可以观察一个人怎样把自己的经验带进 AI,怎样把产出交给别人,又怎样在下一次创作中改变。这给了我一个能够动手做、持续验证的起点。Creators are a starting point because expression, audience, products, income and feedback become tangible here. I can observe how someone brings experience into AI, delivers the result to others, and changes their next piece of work. It is a place to build and keep testing.

我希望逐步找出一种工作方式,让人有能力提出自己的问题、组织自己的创造,并通过真实交付与社会建立联系。人怎样与 AI 结合,又怎样因此获得更自主的位置——这是我愿意持续投入的方向。I want to help discover ways of working in which people can frame their own problems, organize their own creation and connect with society through real contributions. How people work with AI, and gain greater agency through that relationship, is the direction I want to keep pursuing.

DH / SELECTED-FACELIVE

FaceLiveFaceLive

围绕海外创作者,参与人格设定、长对话记忆、知识检索和图文互动,探索本人不在线时如何持续服务粉丝。AI personas for overseas creators. I worked on persona design, conversation memory, knowledge retrieval and text–image interaction, exploring fan services beyond a creator’s available time.

阅读项目复盘Read project reflection
DH / SELECTED-URL-VIDEO

URL 转视频广告URL to video

围绕电商详情页到数字人口播视频,串联商品资料、脚本、配图与视频生成,并继续规划内容矩阵分发。A workflow from product pages to presenter-led videos, connecting product data, scripts, imagery and generation, with further planning for content distribution.

阅读项目复盘Read project reflection
DH / SELECTED-OPEN-BOT

Open BotOpen Bot

参与平台功能框架、知识库与渠道配置交互,以内部创作者业务的真实需求参与调试和体验优化。I contributed to platform structure, knowledge-base and channel configuration, testing and improving the experience against the needs of an internal creator business.

阅读项目复盘Read project reflection
DH / SELECTED-GROWTH

MyyShop / DHgateMyyShop / DHgate

在红人营销场景中,围绕创作者定位、画像、合作匹配和内容辅助做产品探索,关注创作者收入与商家交付效果。Product exploration in creator positioning, profiles, partnership matching and content support, with attention to both creator income and brand outcomes.

查看出海罗盘Open global compass
DH / 更多项目与工具More projects & tools

项目与工具Projects & tools

更多项目与工具More projects & tools

按领域浏览,直接进入对应工具。Browse by field and open a tool directly.

品牌增长Brand & creator growth

把红人营销与 GEO 放在同一条增长链路:前者解决人与内容的触达,后者解决品牌在生成式搜索中的可见与引用。Influencer marketing and GEO: connect brands to people, content and AI search.

影视创作AI content production

围绕短剧、数字人、商品视频和视觉素材,探索从创意判断到音画生产的完整过程。Short videos, digital humans and visual assets, from creative decisions to production.

跨境电商Cross-border commerce

服务中国品牌和跨境业务团队,覆盖品牌表达、商品内容、多平台分发与增长验证。Brand stories, product content, multi-platform publishing and growth validation.

求职招聘Recruiting & careers

覆盖企业招聘协作、候选人体验、职位匹配与 AI 猎头。从早期方案到实际交付,已有六篇项目复盘可查阅。Employer collaboration, candidate experience, job matching and AI recruiters.

量化投研Investment research

围绕信息获取、产业链、个股判断、交易结构和复盘,把研究过程做成可查询、可解释、可验证的工具。Tools for information, industry chains, company research, trading structure and review.

DH / PROJECT-GROWTH

品牌增长Brand & creator growth

把红人营销与 GEO 放在同一条增长链路:前者解决人与内容的触达,后者解决品牌在生成式搜索中的可见与引用。Influencer marketing and GEO: connect brands to people, content and AI search.

相关工具与项目Related tools & projects

继续了解Explore further

项目案例与完整复盘Project cases & full reflections
DH / PROJECT-FILM

影视创作AI content production

围绕短剧、数字人、商品视频和视觉素材,探索从创意判断到音画生产的完整过程。Short videos, digital humans and visual assets, from creative decisions to production.

相关工具与项目Related tools & projects

继续了解Explore further

项目案例与完整复盘Project cases & full reflections
DH / PROJECT-COMMERCE

跨境电商Cross-border commerce

服务中国品牌和跨境业务团队,覆盖品牌表达、商品内容、多平台分发与增长验证。Brand stories, product content, multi-platform publishing and growth validation.

相关工具与项目Related tools & projects

继续了解Explore further

项目案例与完整复盘Project cases & full reflections
DH / PROJECT-CAREER

求职招聘Recruiting & careers

覆盖企业招聘协作、候选人体验、职位匹配与 AI 猎头。从早期方案到实际交付,已有六篇项目复盘可查阅。Employer collaboration, candidate experience, job matching and AI recruiters.

相关工具与项目Related tools & projects

继续了解Explore further

项目案例与完整复盘Project cases & full reflections
DH / PROJECT-RESEARCH

量化投研Investment research

围绕信息获取、产业链、个股判断、交易结构和复盘,把研究过程做成可查询、可解释、可验证的工具。Tools for information, industry chains, company research, trading structure and review.

相关工具与项目Related tools & projects

继续了解Explore further

项目案例与完整复盘Project cases & full reflections
DH / 过往经历Experience

过往经历Experience

按经历展开,看看这些产品选择是怎样一路形成的。The work behind the product choices.

把 AI 接入跨境红人营销。Brought AI into cross-border influencer marketing.

2025 — 2026
在 Myyshop / DHgate,我围绕广告主和海外网红做 AI 产品:出海可行性评估、红人营销方案、品牌出海故事包装、商单生成、网红筛选、商单推荐、网红定位和社媒营销报告。
2025 — 2026
At Myyshop / DHgate, I built AI products around advertisers and overseas creators: go-global feasibility assessment, influencer marketing plans, brand story packaging, brief generation, creator screening, brief recommendation, creator positioning, and social marketing reports.

探索 AI 如何结合自媒体内容副业和变现。Explored how AI can support creator side businesses and monetization.

2024 — 2025
在起点云,我围绕小红书和短视频创作者做工具矩阵:带货数字人、模特换装、商品换背景、AI 短剧解说、红薯助手和种草接单系统。重点是让普通创作者完成内容生产、账号包装和接单变现。
2024 — 2025
At Qidianyun, I built a tool matrix for Xiaohongshu and short-video creators: commerce digital humans, model try-on, product background replacement, AI drama narration, Rednote assistant, and seeding-order systems. The focus was helping everyday creators produce content, package accounts, and monetize orders.

验证电商详情页到数字人口播视频。Validated the path from product pages to talking-head commerce videos.

2024
在来画短暂阶段,围绕 URL to Video 业务做验证:解析电商详情页、清洗素材、优化文案、生成数字人口播视频,并规划内容矩阵分发。
2024
During a short period at Laihua, I validated URL-to-Video workflows: parsing ecommerce pages, cleaning assets, optimizing scripts, generating digital-human talking videos, and planning content distribution.

把网红做成可以持续服务粉丝的 AI 分身。Built AI clones that help influencers serve fans continuously.

2023 — 2024
Facelive 面向海外网红。粉丝需要陪伴、内容、互动和回应,网红本人无法处理所有重复服务。我参与构建人格 Prompt、长对话记忆、RAG 检索、LoRA 形象、主题素材和图文联动推荐。
2023 — 2024
Facelive served overseas influencers. Fans need companionship, content, interaction, and responses, while influencers cannot handle every repetitive service themselves. I worked on persona prompts, long-conversation memory, RAG retrieval, LoRA identity images, themed assets, and image-text recommendations.

把猎头服务拆成 Agent 流程。Turned headhunter services into an Agent workflow.

2023
Jagura Job 面向墨西哥蓝领人群。产品不是简单聊天,而是让 Agent 承接猎头的工作:询问背景、补齐简历、理解偏好、清洗职位、推荐岗位和解释匹配。
2023
Jagura Job served blue-collar workers in Mexico. The product was not just chat; the Agent took on headhunter work: asking about background, completing resumes, understanding preferences, cleaning job data, recommending roles, and explaining matches.

探索 Web3 时代的招聘关系。Explored recruiting relationships in the Web3 era.

2022
在云歌早期招聘方向里,我关注简历确权、个人数据归属、猎头协作和企业职位供给。核心问题是:候选人、猎头和企业能否在一个更透明的网络里共同创造和共享价值。
2022
In Yunge's early recruiting direction, I focused on resume ownership, personal data rights, headhunter collaboration, and job supply. The core question was whether candidates, headhunters, and companies could co-create and share value in a more transparent network.

系统补充用户体验、用户研究和商业数据分析。Added systematic training in UX, user research, and business data analysis.

2021 — 2022
在英国桑德兰大学学习 Applied Business Computing,课程包括 User Experience Design、Research、Software Enterprise、Business Intelligence & Strategy。
2021 — 2022
At the University of Sunderland, I studied Applied Business Computing, including User Experience Design, Research, Software Enterprise, and Business Intelligence & Strategy.

提升企业招聘效率,也改善候选人理解公司的方式。Improved recruiting efficiency for companies and helped candidates understand employers more clearly.

2017 — 2020
在智联招聘参与 OfferKit 和 KA-CRM。OfferKit 关注学生拿到 Offer 后如何持续了解企业、做出选择;KA-CRM 关注销售、售前和执行团队如何协作服务大客户。
2017 — 2020
At Zhaopin, I worked on OfferKit and KA-CRM. OfferKit focused on how students keep learning about employers after receiving offers; KA-CRM focused on collaboration among sales, presales, and delivery teams for key accounts.

学习移动互联应用技术,进入产品和交互的基础阶段。Studied mobile internet application technology and built the foundation for product and interaction work.

2014 — 2017
在深圳信息职业技术学院学习 Android、iOS、UI、Java、H5 和软件测试。这个阶段让我理解一个产品从界面、交互到技术实现之间的基本关系。
2014 — 2017
At Shenzhen Institute of Information Technology, I studied Android, iOS, UI, Java, H5, and software testing. This gave me a basic understanding of how interface, interaction, and implementation connect.

DH / 复杂问题、智能复利
与人的长期选择
Complex problems, compounding intelligence
and long-term choice

Know-how

我如何判断 AI 产品、
个人系统与长期方向?
How I judge AI products,
personal systems and long-term direction.

这些判断不依附于某篇文章。它们来自我对产品实践、个人工作方式与长期方向的持续思考,也需要在真实任务里不断修正。These are working judgments shaped by product practice, personal systems and long-term choices. They stand on their own and remain open to correction through real work.

先选复杂问题,再选择 AI 能力。Choose the complex problem before choosing the AI.

我不再先问“AI 还能做什么功能”,而是先问“谁想完成什么结果”。生成一篇文章、一个素材或一份报告只是 Feature;持续帮助一个人完成增长、研究、决策与复盘,才是值得构建的复杂任务。I no longer begin with what feature AI can perform. I begin with the outcome someone is trying to achieve. An article, asset or report is a feature; sustained help with growth, research, decisions and reflection is a complex task worth building around.

模型强十倍以后,产品会消失还是更强?When models become ten times stronger, does the product disappear—or grow stronger?

这是我判断 AI 产品长期价值的压力测试。如果模型升级后,产品只剩下一层更薄的界面,它大概率只是阶段性 Wrapper;如果模型越强,系统越能承接大问题、理解深层 Context、获得真实反馈,它才可能持续增值。This is my stress test for lasting value. If a model upgrade leaves only a thinner interface, the product is probably a temporary wrapper. If stronger models let the system take on larger problems, understand deeper context and learn from real feedback, it can keep gaining value.

Agent 的壁垒不是功能,而是智能复利。An agent's moat is not its features, but compounding intelligence.

真正不可替代的不是调用了哪个模型,而是系统能否进入真实任务:行动、得到结果、吸收反馈、调整判断,再承接更复杂的工作。任务历史、反馈路径、长期上下文与判断标准连成闭环,智能才会复利。The durable advantage is not which model is called, but whether the system enters real work: acting, observing results, absorbing feedback, changing its judgment and taking on harder tasks. Task history, feedback paths, context and standards must form a loop before intelligence can compound.

Personal OS 的第一职责是熵减。The first job of a Personal OS is to reduce entropy.

个人 AI 系统不是 Skill 和 Agent 的收藏夹。它应该把 Context、Intent、Task、Agent、Result 与 Memory 串成一条可理解的路径,让我管理的东西越来越少,能够解决的问题越来越大。能力可以继续增加,但交互与决策负担必须下降。A personal AI system is not a collection of skills and agents. It should connect context, intent, tasks, agents, results and memory into a legible path—reducing what I manage while increasing the scale of problems I can solve. Capability may grow; interaction and decision load must fall.

Coding 的意义,是把认知变成系统。Coding matters because it turns knowledge into systems.

我学习 Coding,不是为了把“程序员”当成新的职位标签,而是获得认知资产化的能力:把一次理解逐步变成 Framework、Prompt、Skill、Agent、Workflow,最后变成可以执行、验证与迭代的 Product。学习因此不只停在笔记里。I learn to code not to adopt a new job title, but to make knowledge executable: turning an understanding into a framework, prompt, skill, agent, workflow and eventually a product that can be tested and improved. Learning should not stop at notes.

长期方向不是一个赛道,而是一个值得解决的问题。A long-term direction is not a category, but a problem worth solving.

Calling 对我不是抽象的热爱,而是一个我愿意用五到十年投入、承受负反馈并持续校准的重要问题。它需要足够大、足够深,能形成真实反馈,也要有我独特的经验与非共识空间。具体产品可以变化,长期问题不应频繁漂移。Calling is not abstract passion. It is an important problem I am willing to spend five to ten years on, absorbing negative feedback and recalibrating. It must be large and deep enough for real feedback, while leaving room for my experience and a non-consensus view. Products may change; the long-term problem should not drift constantly.

AI 落地,为什么总卡在模型之外?Why does AI adoption get stuck beyond the model?

单点生成可用,整条工作链路仍可能断在上下文、交接和责任上。AI 落地需要重新设计工作:明确输入、判断节点、失败处理和交付标准,让模型能力真正进入业务。Usable generation can still leave a workflow broken at context, handoffs or accountability. Bringing AI into practice requires clear inputs, decision points, failure handling and delivery standards.

未来的公司,还需要按岗位组织吗?Will companies still need to organize around job roles?

当部分执行可以交给 Agent,组织也可以探索围绕问题与交付组队。团队能否更小、更灵活,取决于上下文能否共享、判断能否及时发生、责任能否说清。公司的边界值得重想,信任与协作仍然重要。As agents take on parts of execution, organizations can explore forming around problems and outcomes. Smaller, more flexible teams depend on shared context, timely decisions and clear accountability. Company boundaries may change; trust and collaboration still matter.

数字分身,如何真正延伸一个人?How can a digital persona meaningfully extend a person?

形象和声音只是入口。更值得积累的是一个人的知识、表达、偏好与工作反馈。数字分身需要在授权范围内延伸服务,能被本人纠正,也知道什么时候把判断交还给人。Appearance and voice are entry points. Knowledge, expression, preferences and feedback offer deeper continuity. A digital persona should extend service within authorized boundaries, accept correction and know when to return a judgment to the person.

当执行可以委托,人更需要什么?When execution can be delegated, what matters for people?

提出值得解决的问题,理解真实处境,对结果作判断并承担责任。Human in the loop 应贯穿输入与输出,让人的经验进入协作,也让社会反馈回到下一次创造。Framing worthwhile problems, understanding real situations, judging results and taking responsibility. Human involvement should span input and output, bringing experience into collaboration and feedback into the next act of creation.

DH / 红人营销系统Creator marketing system

红人营销系统Creator marketing system

连接创作者与商家,围绕定位、匹配和内容辅助,探索更可靠的商业合作。Connecting creators and brands through positioning, matching and content support.

先理解合作双方Understand both sides

我在 MyyShop / DHgate 的产品探索,围绕创作者定位、画像、注册引导、合作匹配和内容辅助展开。创作者需要收入机会,商家需要符合目标的内容与交付。At MyyShop / DHgate, my product work explored creator positioning, profiles, onboarding, matching and content support. Creators need opportunities; brands need relevant content and dependable delivery.

把匹配接到真实交付Connect matching to delivery

我关注的不只是推荐出一个账号,还包括商家目标是否明确、创作者是否适合、内容要求能否被理解,以及合作之后发生了什么。方案推进也受到数据、预算和产品入口的限制。I considered brand goals, creator fit, content requirements and what happened after a match. Data, budget and product-entry constraints also shaped what could be taken forward.

DH / 种草系统Seeding & collaboration

种草系统Seeding & collaboration

在起点云的创作者业务中,参与衔接内容创作与接单变现的种草系统。At Qidianyun, I worked on connecting creator content production with paid collaboration.

从创作延伸到接单From creation to collaboration

这项工作属于起点云的创作者副业产品方向,重点是让内容创作与种草接单产生连接。相比单独增加一个生成工具,我更关注内容做出来之后,能不能进入实际合作。This work belonged to Qidianyun’s creator tools portfolio. It connected content creation with seeding collaborations, focusing on what happens after content is produced.

我持续关心的问题The question that stayed with me

创作者能不能接到适合自己的单,商家能不能收到符合需求的内容?这也延续到了我后来对红人营销双边匹配和交付的思考。Can creators find suitable work, and can brands receive content that meets their needs? That question carried into my later work on creator-brand matching and delivery.

DH / 漫影助手Animation assistant

漫影助手Animation assistant

围绕二维动画创作,做过市场判断、产品规划和 AI 场景、角色、动作与表情的技术选型探索。Product planning and technical exploration for 2D animation, including AI scenes, characters, motion and expression.

从具体制作场景开始Start with a production scene

规划从基础动画工具切入:场景、道具、音频、图层、字幕和角色表情,再考虑用 AI 降低部分素材制作的门槛。The plan began with scenes, props, audio, layers, subtitles and expressions, then considered AI for parts of asset production.

做过的方案探索Design explorations

记录包括 LoRA 场景生成、角色生成、动作与表情控制,以及动态表情包的组合流程。这部分属于历史规划和技术方案,没有把候选能力当作已上线功能。Records include LoRA scene generation, character creation, motion and expression control, and a workflow for animated stickers. These were plans and technical proposals, not a claim that every candidate feature shipped.

DH / 短剧解说助手Drama commentary assistant

短剧解说助手Drama commentary assistant

围绕推剧助手,探索把视频与字幕转成分镜、剧情概要和解说文案,并参与声音与画面相关实验。Work on Tuiju Assistant explored turning video and subtitles into storyboards, plot summaries and commentary, alongside voice and visual experiments.

把故事上下文带进解说Give commentary its story context

我做过字幕、分镜、概要与解说的提示词和流程探索。不同阶段的输入输出并不天然兼容:时码、对白和故事背景如何传递,会直接影响下一步的理解。I explored prompts and workflows for subtitles, storyboards, summaries and commentary. Timecodes, dialogue and story context need to survive the transitions between stages.

生成之后,还要能用Generation still needs to be usable

历史记录里,解说效果仍是待解决的问题,也涉及下载、二创和发布引导的缺口。这段工作让我更关注完整制作链路,而不只是一段能生成文字的提示词。Commentary quality remained an open problem in the records, alongside gaps in download, editing and publishing guidance. This work drew my attention to the full production process.

DH / 红薯助手Xiaohongshu assistant

红薯助手Xiaohongshu assistant

小红书内容商业化辅助方案,把账号定位、需求研究、内容拆解与商单协作放到同一条工作流里。A Xiaohongshu product design connecting account positioning, audience research, content analysis and brand collaborations.

定位决定后面的内容Positioning comes before content

方案区分博主人设号与商家商业号,从账号定位、真实用户问题、选题和内容结构出发,避免只靠泛流量衡量效果。The design distinguished creator-led and merchant accounts, starting from positioning, real audience questions and content structure rather than broad traffic alone.

让 Brief 成为生成的依据Ground generation in the brief

留下了品牌 Brief 字段、账号与产品输入、自我介绍和合作构思提示词,以及排版与交付检查点。人设、受众和产品事实不能为了文案通顺而被模型改写。这些是设计与调研阶段的工作资产。The work includes brand-brief fields, account and product inputs, introduction and collaboration prompts, and delivery checks. The model must not rewrite identity, audience or product facts to make copy sound better. These are design and research assets.

DH / 带货助手Commerce content assistant

带货助手Commerce content assistant

承接过带货相关需求,围绕带货场景文案和商品内容开展产品工作。I took on commerce-content requirements, including copywriting and product-content work.

从商品表达进入创作Start with how the product is presented

历史会议记录包含带货场景文案工具和需求梳理;同阶段的职责还涉及带货数字人、模特换装和商品换背景等 AI 内容方向。Historical meeting notes cover commerce-copy tooling and requirements. Work in the same period also involved digital presenters, virtual outfit changes and product-background generation.

把工具放回使用场景Keep the use case in view

这些尝试面向缺少拍摄资源的创作者。相关技术方向不等于都归入同一个版本;这里保留具体工作范围,不把工具组合写成已验证的收入结果。These explorations addressed creators with limited filming resources. The technical directions did not necessarily belong to one release, and the work is not presented as a validated income outcome.

DH / FaceLiveFaceLive

FaceLiveFaceLive

围绕海外创作者,探索用 AI 分身延伸本人不在线时的粉丝服务。AI personas for overseas creators, exploring fan services beyond a creator’s available hours.

围绕一个具体的人Build around a real person

参与人格设定、长对话记忆、知识检索和图文互动,让创作者的表达、形象与粉丝关系进入 AI 服务。I contributed to persona design, conversation memory, knowledge retrieval and text-image interaction around a creator’s voice and audience relationships.

从互动走向收入From interaction to income

这段经历包含早期互动付费探索,让我开始思考 AI 如何延伸个人服务。局部付费并不等于商业模式已经能够稳定复制。Early paid-interaction experiments made me consider how AI could extend personal services. Local payment validation did not establish a repeatable business model.

阅读完整项目复盘Read the full project reflection ↗