[DIS] AI Agent 驱动的内容打赏与策展系统 | AI Agent-driven Content Tipping and Curation System

TLDR:我们申请 81,000U(约 18.6M CKB)用于开发一个基于 AI Agent 的内容策展平台。该平台将整合 Fiber 闪电网络的微支付能力,通过 AI 自动评估内容质量并实时打赏优质创作者,旨在构建一个去中心化、高效且公平的内容生态系统。

各位 CKB 社区的伙伴们,你们好!我很荣幸向社区阐述我们的提案。

我们计划将 AI Agent 与 Fiber 闪电网络结合,构建一个去中心化的内容激励策展 Agent,使其成为自动化的「策展人 + 打赏者」。

通过这个 AI Agent 驱动的打赏策展系统,我们将构建一个去中心化、高效自动的内容生态。系统利用 AI 评估内容质量,并通过 Fiber 闪电网络为优质创作者提供低成本、高频的实时打赏。这不仅让创作者获得更公平的收益,同时也为 CKB 生态带来大规模应用场景。

我们致力于打造一个人人公平参与、价值直接流通的内容分发网络,让每个优质内容都能被发现。通过结合 AI 智能化与区块链去中心化特性,我们将重塑内容的创作、分发与变现模式。

一、背景与机遇

(一)AI Agent 的发展历程与价值证明

1. AI Agent 的技术发展现状

  • 已从概念验证阶段步入实用化阶段,具备强大的 LLM 基础能力和多模态处理能力,目前正处于快速发展期
  • 2024 年实现突破性创新,2025 年正朝向专业化和垂直领域应用的方向发展
  • 技术能力大幅提升,可支持复杂的实际应用场景

2. 当前阶段进行 AI Agent 应用开发的优势

  • 市场需求明确:Web3 市场度过第一波 Agent 热潮后,市场更关注具有实际应用价值的 Agent 产品。在投机炒作退潮后,虽然大多数项目陷入沉寂,但像 AIXBT 这类专注于加密投资决策和市场行情分析的 AI Agent 应用却持续获得用户增长,显示出用户和市场对实用型 AI Agent 的认可度不断提升。
  • 应用场景丰富:从B端(如金融、医疗、法律等专业领域)到C端(如游戏、娱乐、教育等创意领域)都有广阔空间
  • 基础设施成熟:去中心化计算、可验证计算(TEE、ZKP)等基础设施日趋完善,为智能体提供了安全可靠的运行环境
  • 微支付能力具备:Fiber 闪电网络和链上稳定币(如 USDT)已实现近零成本和亚秒级结算,解决了传统区块链在高频小额支付场景中的瓶颈,闪电网络也即将迎来 USDT 等稳定币产品

3. AI Agent 具备独特的价值主张和应用场景

  • 自动化与高效性:自动筛选、分析和执行任务,极大提升效率,减少人工干预,适合处理海量数据和复杂场景
  • 去中心化与透明激励:结合链上技术,AI Agent的决策和奖励机制可实现公开透明和可审计,显著提升公平性。通过去中心化技术,AI Agent能构建自动化、透明且公平的经济激励闭环
  • 多样化和专业化:智能体在投资、娱乐、DeFi、游戏、社交、Dao 治理等领域均有突破,能够满足不同细分市场的需求,形成差异化竞争优势
  • 创新应用模式:开创新的商业模式,如自主交易、内容创作等

4. 为什么现在是落地 AI Agent 的最佳时机

  • 正处于从"技术萌芽"到"商业落地"的关键转折期。AI Agent 的技术与应用正快速走向成熟和专业化,市场潜力巨大。
  • 市场和文化认知日趋成熟,用户接受度不断提升。首波 AI Agent 热潮退去后,市场对真正实用的AI Agent 落地产品形成了更理性的期待。随着基础设施、经济模型与市场需求同步成熟,现在正是抢占先机、塑造未来智能体生态的最佳时机。

(二)CKB 的战略性契合点

未来为 AI 服务的支付体系须具备以下五大核心能力:

  • 微支付能力:AI 智能体的交易涉及极小金额(如 $0.0001)且高频,因此手续费必须极低,接近于零。
  • 毫秒级交易结算:AI 智能体发生在亚秒级时间内,支付系统必须能够实时结算,而非等待数分钟或数天。
  • 去中心化与抗审查:AI 智能体需要自主交易,不依赖中心化金融机构。
  • 全球可用性:AI 智能体不受地域限制,支付体系必须支持跨境交易,无需依赖银行账户。
  • 智能化支付协议:支付系统必须与 AI 智能体无缝交互,支持自动结算、智能交易路由和流动性优化。

将 AI Agent 与 CKB/Fiber 结合开发内容打赏策展系统具有以下关键优势:

1. 技术契合度高

  • CKB 的 Fiber 闪电网络完美满足 AI Agent 的支付需求,提供近零成本、抗审查的支付通道、毫秒级结算、去中心化特性和全球可用性。它是目前仅有的能够满足 Agent 时代支付要求,同时兼顾用户体验和扩展性的基础设施
  • 交易支付性能可以比肩传统 Web2 支付体系,性能足以支撑大规模自动化内容打赏
  • 基于 CKB 的 RGB++ 协议,可完美支持 sats、USDT 和各类社区主流币种的打赏与支付
  • 策展内容的后续玩法可充分利用 CKB 灵活的 Cell 模型,基于此还可与 Nostr Protocol 去中心化的社交协议进行绑定映射,比其他公链具有更强的创新空间

2. 生态协同效应

  • 为 Fiber 闪电网络提供大规模实际应用场景,有效促进基础设施发展
  • 目前 CKB 生态中尚未出现同类 AI Agent 项目,具备明显的先发优势
  • AI Agents 将成为链上生态的重要参与主体,为 CKB 生态发展注入新动力

(三)为什么从打赏和策展切入 AI Agent 应用?

当前内容创作生态面临以下关键痛点:

  • 中心化垄断:大多数优质内容的曝光完全由平台算法控制,存在黑箱操作且过分追求商业利益,致使中长尾创作者无法获得公平的曝光机会
  • 激励机制僵化:人工策展效率低下且主观性强,无法有效应对海量内容的筛选需求

AI Agent 结合自动化、可验证性(通过 ZK 证明)、实时反馈和去中心化激励等优势,通过 Fiber 闪电网络实现即时微支付功能。

通过 AI Agent 自动化策展和打赏流程,我们提供以下解决方案:

  • 效率提升:基于 AI 多模态评估系统(信息熵、原创性、社区价值),实现毫秒级内容评估与打赏,显著提升策展效率
  • 成本降低:利用 Fiber 闪电网络,将单次策展打赏成本大幅降低至极小金额,显著优于传统平台的手续费水平
  • 智能策展:从技术文档到艺术作品,AI 能评估多样化内容的价值,并利用 zkML 技术把评估过程转化为可验证的数学证明,从而消除黑箱操作
  • 价值回归:打破平台垄断,使创作者直接获得作品价值回报,构建更公平透明的内容生态

这一创新机制不仅解决内容生态痛点,还为 Fiber 闪电网络提供大规模应用场景,实现技术创新与市场需求的结合。随着 AI 技术进步和 Web3 基础设施完善,我们将打造更开放、公平、高效的内容创作生态系统。AI Agent 策展系统将提升创作者和消费者的参与度,促进生态良性发展。

二、机制设计

(一)双重角色 Agent

  1. 策展人(Curator)AI 对用户生成内容(UGC)进行分析,从海量数据中筛选优质内容,主要从信息密度、原创性和社区共鸣度三个维度进行评估。

  2. 打赏者(Tipper)通过 Fiber 网络为优质内容提供自动微额打赏(如 10-100 sats 或未来的稳定币)。打赏资金来源包括:1. 协议金库(初期补贴);2. 社区打赏广告分成(后期可持续);3. 用户付费策展内容。打赏标准、额度范围和策展标准将在实际运营中持续优化。

(二)机制流程示意

(三)创作者激励

  • 评价标准简单明确:只考量内容质量。无论是大 V、KOL、普通用户还是新手,AI Agent 都采用统一标准进行内容激励和策展。评分与打赏规则完全透明且可审计,相比 Kaito 等依赖流量、注意力变现的产品,这更能让中长尾用户受益。
  • 优质内容创作的激励机制将为市场带来积极影响,提高创作者的参与热情。

(四)扩展玩法

1、用户主动打赏

用户可在赞助或充值后,指示 Agent 对特定内容进行打赏,培养用户使用 Agent 进行内容打赏的习惯。

例:用户可在推文下@Agent 进行打赏,Agent 将以用户身份完成打赏并评论。

2、主动策展

用户可付费请求特定主题的内容策展。

例:用户可要求 Agent 策展"新手如何快速学习上手 MCP"或"链上交易经验分享"等主题,Agent 会加大对相关主题的激励力度,促进用户创作。

(五)技术风险与解决方案

风险 解决方案
女巫攻击(为打赏而刷内容) 严格检测内容相似度,限制用户行为和打赏频次
Fiber 闪电网络可用性和易用性问题 等待 Joyid 集成 Fiber 闪电网络和离线支付功能,通过 UTXOStack 等应用落地进一步优化新用户注资等用户体验问题
AI 评估偏差 初期采用 Prompt 提示词工程,后续引入模型微调和人工策展校正机制

三、实施路径

项目计划分三个阶段推进,实际执行将根据进展情况灵活调整:

阶段 核心目标 关键工作 具体任务
Alpha阶段 基于 X 平台完成 Agent 应用的开发与上线 1. 团队筹备
2. 技术开发
3. 运营准备
核心基础框架搭建
AI 策展系统开发
打赏通道对接
基于 X 平台的内容检索引擎
应用冷启动方案细化与执行
Beta阶段 产品功能迭代完善和扩大用户与影响力规模 1. 上线扩展玩法
2. 迭代优化使用体验
3. 市场拓展与增长
功能完善与体验优化
强化安全检测机制
运营推广主动打赏和主动策展玩法
达成既定用户增长和合作伙伴目标
正式运营阶段 打造成 Web3 Agent 领域的标杆应用 1. 持续扩大用户规模和影响
2. 落地扩展 Agent 应用到更多平台
尽可能覆盖全面 Web3 领域内容【全面覆盖存在技术挑战】
推进生态建设和升级商业模式
封装推广 Agent 的打赏能力

1. Alpha 阶段(1~2个月)

完成基础框架搭建,包括 AI 策展评估系统和 Fiber 打赏通道对接,优先在 X 平台的 CKB 生态内容领域验证核心机制

  • 完善 AI Agent 策展评估系统,完成 X 平台对接,构建 X 平台内容检索引擎
  • 基于 X 平台,完整实现 AI Agent 策展和打赏的核心流程机制

风险点:

  • 目前 Joyid 对 Fiber Network 的支持仍在开发测试中,若期间未能完成,将暂时使用 CKB 链打赏方式替代
  • 被打赏用户需提供接收地址,这种早期体验较为繁琐,可能影响打赏转化率。我们将重点优化 X平台上的用户地址绑定服务

冷启动策略

  • 适度降低 Agent 评估标准,提升早期用户体验
  • 猎手奖励:用户推荐未发掘的优质内容,若后续被 AI 打赏,推荐者获10%分成
  • 策展游戏化:每周发布「策展任务」(如「量化交易最佳入门教程」),优胜者获得任务奖励

2. Beta 阶段(2~3个月)

扩大策展和打赏的内容领域,提升影响力和用户规模,完善产品功能和用户体验

  • 完善防范女巫攻击和恶意刷打赏的识别机制,确保 Agent 良性运转
  • 实现用户主动打赏、主动策展等扩展功能,推广至 X 平台用户群,打造标杆效应
  • 接入 Fiber 闪电网络打赏,优化打赏流程和用户体验
  • 与项目方、社区合作开展特定领域内容策展,扩大用户基础和打赏资金池
  • 针对 Web3 KOL 群体进行定向打赏和策展,扩大项目影响力

3. 正式运营(长期)

全面覆盖 Web3 和 Crypto 领域的优质内容打赏与策展,围绕用户真实需求,将其打造成融入日常场景的实用 AI Agent 应用

  • 持续运营推广,扩大影响力和用户规模,让"打赏"、"策展"深入用户心智,这是未来工作重点
  • 通过打赏内容曝光吸引 Sponsor 对打赏池进行赞助
  • 将 Agent 服务扩展至 Discord、Telegram 等平台,拓展用户群和影响力
  • 将打赏能力封装供其他用户和Agent应用调用,在打赏领域建立 Agent 生态地位
  • 紧跟 AI、AI Agent 技术发展,融入 Web3 社区创新玩法,持续推进

四、预算申请与分配

本提案计划向 CKB Community Fund DAO 申请总计 81,000U,按提案编写时 CKB 价格 0.004348 计算,折合 18,629,254 CKB

1. 预算详情

类别 预算 说明
人力补贴成本 36000U 预估1名全职开发+1名全职运营,每人每月 1500U x 12 个月,兼职则按 50% 计算
开发成本 20000U 包含常规后端服务部署、 X 平台开发账号、大模型 Token 消耗、和因开发、服务、运营所需的第三方服务等成本
运营成本 25000U 其中 10000U 用于 Agent 打赏池,剩余用于冷启动期、活动推广期及长期运营推广

所有预算类别将在一年或更长时期内逐步分配使用,具体如下:

1. 人力补贴成本(36000U):

  • 每月每人补贴 1500U,覆盖 1 名全职开发和 1 名全职运营,按月线性发放(12 个月补贴期)
  • 若团队成员为兼职投入,则补贴减半(750U/月)

2. 开发成本(20000U):

  • 【20%】后端应用部署服务器、数据库等费用
  • 【45%】X 平台开发账号费用【基于 X 平台的官方服务及延伸的内容引擎服务成本比较高】
  • 【15%】AI 大模型 API 调用费用
  • 【10%】第三方服务订阅费用(如监控、数据分析等)

3. 运营成本(25000U):

  • 【40%】Agent 打赏池资金:10000U 【作为从预算中投入打赏池的最高额度,需在耗尽前实现开源增收】
  • 剩余 15000U 分配用于:
    • 【6%】冷启动期推广活动
    • 【9%】活动期用户激励
    • 【45%】一年以至更长期的内容营销、用户激励活动、社区运营等费用

团队承诺会详细记录并公开所有资金使用明细,确保资金使用透明有效。

2. 里程碑与进度考核机制

为确保项目稳步推进和资金高效使用,我们设置了以下分期里程碑机制和对应的考核指标:

阶段 发放条件 具体金额 预计时间周期 CKB 折算
Milestone0 提案通过后发放,用于项目启动 10000U 提案通过后 2299908
Milestone1 完成基础框架搭建,实现核心功能,在 X 平台正式上线并落地打赏和策展能力 13000U 1-2个月 2989880
Milestone2 完成上线用户主动打赏、主动策展扩展等功能玩法,累积 X 平台 500+ 关注用户 13000U 2-3个月 2989880
Milestone3 累积 X 平台 1.8K+ 关注用户,真实用户使用主动打赏能力 10次+,真实策展打赏 500 条+内容 14000U 1~2 个月 3219871
Milestone4 累积 X 平台 5K+ 关注用户,进行 1K+ 条真实内容打赏,真实用户使用主动打赏能力 100次+,接受 3个+ Sponsor 赞助 15000U 2个月左右 3449862
Milestone5 累积 X 平台 10K+ 关注用户,打赏池增收 3000U+ 16000U 2~3个月 3679853

启动阶段(Milestone 0)

初始资金 10000U 用于基础开发环境搭建和团队组建。这笔资金将在提案通过后立即发放,以支持项目快速启动,推动核心功能开发落地。

基础框架阶段(Milestone 1)

完成核心功能开发并在 X 平台实现首次部署。具体考核指标:

  • 实现 AI Agent 的内容策展与评估系统,完成与 X 平台的对接,搭建 X 平台的内容检索引擎
  • 在 X 平台上成功实现 AI Agent 的策展与打赏全流程运转
  • 达成后发放 13000U

功能扩展阶段(Milestone 2)

丰富产品功能并验证市场接受度。考核指标:

  • 完成用户主动打赏和主动策展功能的开发与部署
  • 完成防范女巫攻击的安全机制,优化用户打赏流程的体验
  • 集成并跑通上线基于 Fiber 闪电网络打赏流程
  • 达到 500+ 真实用户关注,完成 80+ 次内容打赏
  • 达成后发放 13000U

补充说明

我们深知里程碑的设定和明确的考核标准对社区评审提案和监督资金使用的重要性。与以往的应用型提案不同,过往通过的提案主要关注功能明确的应用开发与落地,可以简单按照功能模块设置里程碑进行考核,较少涉及产品落地后的使用情况。而我们的 Agent 应用周期较长,计划按一年时间分阶段使用预算,有序推进。虽然前期里程碑 1 和 2 侧重技术实现和交付,但从长远来看,持续的运营和迭代能力才是关键,市场认可度和用户欢迎程度是我们的终极评价指标。

后续三个里程碑暂将主要通过可量化的运营数据指标来考核。我们当然会持续进行技术迭代和平台扩展,但具体在哪个阶段实施需要根据产品和市场情况决定,因此暂未列入考核指标。我们设定的运营考核指标虽不算激进,但能确保在项目稳步推进。如果社区对里程碑和考核指标有更合理和建设性的建议,我们非常欢迎在提案讨论阶段提出。

市场验证阶段(Milestone 3)

验证产品市场匹配度。考核指标:

  • 累计 1800+ 真实用户关注
  • 用户主动使用打赏功能达到 10 次以上
  • AI 策展系统完成 500+ 次有效内容打赏
  • 达成后发放 14000U

规模增长阶段(Milestone 4)

实现用户规模突破和商业化。考核指标:

  • 累计 5000+ 真实用户关注
  • 完成 1000+ 次内容打赏
  • 获得至少 3 个个人/机构/项目方赞助
  • 达成后发放 15000U

生态建设阶段(Milestone 5)

建立可持续发展模式。考核指标:

  • 累计 10000+ 用户规模
  • 打赏池实现自主增长 3000U+
  • 达成后发放 16000U

阶段衔接和风险管理

后续运营效果受市场行情和 Agent 应用发展热度影响,存在不确定性。如未能按期达到目标,我们将:
• 加强社区反馈机制,适当延长里程碑期限
• 调整用户增长路径策略,确保最终目标不变

若市场验证表明当前阶段的 Agent 应用未能满足有效需求,未达到预期的规模和效果,我们将终止后续预算申请并退回剩余资金。

五、其他说明

我是谁,为什么由我来做这个?

我是一名独立开发者,曾在360、蚂蚁等互联网公司任职。在上一轮市场周期中,我参与开发了某 NFT 租赁协议,并负责某 NFT 项目的发行和运营工作。我熟悉区块链技术和 Web3 生态,坚信区块链的长期价值,并持续专注于 Web3/Crypto 领域。

这个提案的灵感来自于我报名 Rock Web5 黑客松后的一个想法,希望将 AI Agent 与打赏机制相结合。在黑客松期间,我完成了基于 Nostr Protocol 的极简策展与打赏 MVP。经过与伙伴们的深入讨论以及项目方的帮助与反馈后,进行多轮完善,最终形成了现在的版本。

如果本提案能获得 CKB 社区的认可,我承诺将以实名身份在社交媒体上开发,定期公布项目进展和预算使用情况,全力推动项目发展。如果提案未获通过,我也完全尊重社区的决定。欢迎CKB社区成员对本提案的各个方面提出宝贵意见!


English Version


[DIS] AI Agent-driven Content Tipping and Curation System

TLDR: We are applying for 81,000U (~ 18.6M CKB) to develop an AI Agent-based content curation platform. This platform will integrate Fiber’s lightning network micropayment capabilities, automatically assess content quality through AI, and reward high-quality creators in real-time. Our goal is to build a decentralized, efficient, and fair content ecosystem.

Dear CKB community members, greetings! I am honored to present our proposal to the community.

We plan to integrate AI Agent with the Lightning Network of Fiber to construct a decentralized content incentive curation Agent, functioning as an automated “curator + tipper.”

Through this AI Agent-powered tipping curation system, we will build a decentralized, highly efficient, and automated content ecosystem. The system leverages AI to assess content quality and provides low-cost, high-frequency, real-time tips to top creators via the Fiber lightning network. This not only ensures creators receive fairer compensation but also brings large-scale application scenarios to the CKB ecosystem.

We are committed to building a content distribution network where everyone can participate fairly and value flows directly, ensuring that every high-quality content gets discovered. By integrating AI intelligence with blockchain’s decentralized nature, we aim to reshape the models of content creation, distribution, and monetization.

1. Background and Opportunities

(1) Development History and Value Demonstration of AI Agents

1. The Current State of AI Agent Technology Development

  • It has transitioned from the proof-of-concept phase to practical implementation, boasting robust foundational capabilities in large language models (LLM) and multimodal processing. Currently, it is in a phase of rapid development.
  • Achieving groundbreaking innovations by 2024, and moving towards specialization and vertical application domains by 2025.
  • Significant advancements have been made in technical capabilities, enabling support for complex real-world application scenarios.

2. Advantages of Developing AI Agent Applications at the Current Stage

  • Market demand has crystallized: After the first wave of enthusiasm for Agents in the Web3 market, the market is now focusing more on Agent products with practical application value. Despite the fact that most projects have fallen silent after the speculative hype has subsided, applications like AIXBT, which specialize in crypto investment decision-making and market trend analysis, continue to see user growth. This indicates that the recognition of practical AI Agents by both users and the market is steadily increasing.
  • Rich application scenarios: From B-end (such as finance, healthcare, legal, and other professional fields) to C-end (such as gaming, entertainment, education, and other creative fields), there is ample room for growth.
  • Infrastructure Maturity: Decentralized computing and verifiable computing (TEE, ZKP) infrastructures are becoming increasingly robust, providing secure and reliable operating environments for intelligent agents.
  • Micro-payment capabilities enabled: Fiber’s Lightning Network and on-chain stablecoins (e.g., USDT) have achieved near-zero costs and sub-second settlement, overcoming the traditional blockchain’s limitations in high-frequency, small-value payment scenarios. The Lightning Network is also on the verge of welcoming stablecoin products like USDT.

3. AI Agent has unique value propositions and application scenarios.

  • Automation and Efficiency: Automatic screening, analysis, and task execution significantly enhance efficiency, reduce manual intervention, and are suitable for handling massive data and complex scenarios.
  • Decentralization and Transparent Incentives: By integrating on-chain technology, the decision-making and reward mechanisms of AI Agents can achieve openness, transparency, and auditability, significantly enhancing fairness. Through decentralized technology, AI Agents can construct an automated, transparent, and fair economic incentive loop.
  • Diversification and Specialization: Agents have made breakthroughs in various fields such as investment, entertainment, DeFi, gaming, social networking, and Dao governance, catering to the needs of different niche markets and forming differentiated competitive advantages.
  • Innovative Application Models: Pioneering new business models, such as autonomous trading and content creation.

4. Why Now is the Optimal Time to Implement AI Agents

  • The AI Agent technology and applications are rapidly maturing and becoming more specialized, marking a critical turning point from “technological infancy” to “commercial implementation.” The market potential is immense.
  • As the market and cultural perceptions continue to mature, user acceptance is on the rise. Following the initial wave of enthusiasm for AI Agents, the market now holds a more rational expectation for genuinely practical AI Agent products that can be effectively implemented. With infrastructure, economic models, and market demands all reaching a level of maturity in sync, the present moment represents the optimal time to seize the opportunity and shape the future ecosystem of intelligent entities.

(2) Strategic Fit of CKB

The payment system for AI in the future must possess the following five core capabilities:

  • Micro-payment Capability: Transactions involving AI agents involve extremely small amounts (e.g., $0.0001) and occur at high frequencies, so transaction fees must be minimal, approaching zero.
  • Millisecond-level transaction settlement: AI agents operate within sub-second timeframes, requiring payment systems to settle in real-time rather than waiting for minutes or days.
  • Decentralization and Censorship Resistance: AI agents need to conduct autonomous transactions without relying on centralized financial institutions.
  • Global Accessibility: AI agents are not bound by geographical limitations, and the payment system must support cross-border transactions without relying on bank accounts.
  • Intelligent Payment Protocol: Payment systems must seamlessly interface with AI agents, enabling automated settlement, intelligent transaction routing, and liquidity optimization.

Integrating AI Agent with CKB/Fiber to develop a content tipping curation system offers several key advantages:

High technical compatibility

  • CKB’s Fiber Lightning Network perfectly meets the payment needs of AI Agents, offering near-zero cost, censorship-resistant payment channels, millisecond-level settlement, decentralized features, and global availability. It is currently one of the few infrastructures capable of fulfilling the payment requirements of the Agent era while also balancing user experience and scalability.
  • Transaction payment performance can rival traditional Web2 payment systems, with sufficient performance to support large-scale automated content tipping.
  • The RGB++ protocol based on CKB perfectly supports tipping and payment for sats, USDT, and various community-dominant cryptocurrencies.
  • The follow-up play of curated content can fully leverage the flexible Cell model of CKB, and based on this, it can also be bound and mapped with the decentralized social protocol of Nostr Protocol, offering stronger innovation potential compared to other public blockchains.

2. Ecological Synergy

  • To provide Fiber’s Lightning Network with large-scale practical application scenarios, effectively promoting the development of infrastructure.
  • Currently, there are no similar AI Agent projects in the CKB ecosystem, giving it a significant first-mover advantage.
  • AI Agents will become key participants in the on-chain ecosystem, injecting new vitality into the development of the CKB ecosystem.

(3) Why Focus on AI Agent Applications from the Perspectives of Tipping and Curation?

The current content creation ecosystem faces the following key pain points:

  • Centralized Monopoly: The exposure of most high-quality content is entirely controlled by platform algorithms, which involve opaque operations and excessive pursuit of commercial interests, resulting in mid-to-long tail creators being unable to obtain fair exposure opportunities.
  • Incentive Mechanism Rigidity: Manual Curation is Inefficient and Highly Subjective, Unable to Effectively Address the Need for Screening Massive Amounts of Content.

AI Agent combines the advantages of automation, verifiability (through ZK proofs), real-time feedback, and decentralized incentives, enabling instant micropayments via the Fiber Lightning Network.

Through AI Agent automation of curation and tipping processes, we offer the following solutions:

  • Efficiency Improvement: Leveraging the AI-based Multimodal Evaluation System (Information Entropy, Originality, Community Value), content evaluation and rewards are achieved in milliseconds, significantly enhancing curation efficiency.
  • Cost Reduction: By leveraging the Fiber Lightning Network, the cost per curation reward is significantly reduced to a minimal amount, outperforming traditional platforms’ transaction fees.
  • Smart Curation: From technical documents to artworks, AI can assess the value of diverse content and use zkML technology to transform the assessment process into verifiable mathematical proofs, thus eliminating the black box operation.
  • Value Return: Breaking Platform Monopolies, Enabling Creators to Directly Reap the Rewards of Their Work, and Building a More Fair and Transparent Content Ecosystem.

This innovative mechanism not only addresses the pain points of the content ecosystem but also provides Fiber’s Lightning Network with large-scale application scenarios, achieving the integration of technological innovation and market demand. With the advancement of AI technology and the improvement of Web3 infrastructure, we will build a more open, fair, and efficient content creation ecosystem. The AI Agent curation system will enhance the engagement of creators and consumers, promoting the healthy development of the ecosystem.

2. Mechanism Design

(1) Dual Role Agent

  1. Curator AI analyzes user-generated content (UGC), sifting through vast amounts of data to select high-quality content, primarily evaluated based on three dimensions: information density, originality, and community resonance.
  2. Tippers Automatically reward high-quality content through the Fiber network with micropayments (e.g., 10-100 sats or future stablecoins). Funding sources include:
    1. Protocol treasury (initial subsidies);
    2. Community tip ad revenue sharing (sustainable in the long term);
    3. User-paid curation of content.

Reward criteria, amount ranges, and curation standards will be continuously optimized during actual operations.

(2) Mechanism and Process Diagram

(3) Creator Incentives

  • The evaluation criteria are simple and clear: content quality is the only consideration. Whether you’re a big V, KOL, average user, or newcomer, AI Agent applies the same standard for content incentives and curation. The scoring and tipping rules are entirely transparent and auditable, benefiting mid-to-long-tail users more compared to products like Kaito that rely on traffic and attention for monetization.
  • The incentive mechanism for high-quality content creation is expected to have a positive impact on the market, boosting creators’ enthusiasm for participation.

(4) Extended Features

1. User Initiated Tip

After sponsoring or topping up, users can instruct the Agent to tip specific content, fostering the habit of using the Agent for content tipping.

Users can @Agent under a tweet to tip, and Agent will complete the tip and comment on behalf of the user.

2. Proactive Curation

Users can pay to request curated content on specific topics.

Users can request the Agent to curate topics such as “How Newbies Can Quickly Learn and Get Started with MCP” or “Sharing Experiences in On-Chain Transactions.” The Agent will increase incentives for related topics to encourage user-generated content.

(5) Technical Risks and Solutions

Risk Solution
Witch Attack (Spamming for Tips) Strictly monitor content similarity, limit user behavior, and regulate the frequency of rewards.
Fiber: Issues with Lightning Network’s Availability and Usability Waiting for Joyid to integrate Fiber Lightning Network and offline payment features, further optimizing user experience issues such as new user funding through applications like UTXOStack.
AI Bias Assessment Initially utilizing Prompt engineering, subsequently incorporating model fine-tuning and human curation correction mechanisms.

3. Implementation Path

The project plan will be advanced in three phases, with actual execution adjusted flexibly based on progress.

Stage Core Objectives Key tasks Specific Tasks
Alpha stage Complete the development and launch of Agent application based on X platform 1. team preparation
2. technology development
3. operation preparation
Construction of core infrastructure framework
Development of AI curatorial system
Docking of reward channel
Content retrieval engine based on X platform
Refinement and implementation of application cold-launch program
Beta stage Iterative improvement of product functions and expansion of users and influence scale 1. online expansion of gameplay
2. iteration and optimization of user experience
3. market expansion and growth
Function improvement and experience optimization
Strengthen security detection mechanism
Operation and promotion of active reward and active curation gameplay
Reach the established user growth and partner goals
Formal Operation Stage Become a benchmark application in the field of Web3 Agent 1. continuously expand user scale and influence
2. expand Agent application to more platforms on the ground
Cover as much of the Web3 domain as possible [There are technical challenges in comprehensive coverage]
Promote the ecological construction and upgrade the business model
Encapsulate and promote the Agent’s tipping ability.

1. Alpha Phase (1-2 months):

Completed the basic framework construction, including the AI curation evaluation system and Fiber tipping channel integration, prioritizing the verification of core mechanisms in the CKB ecosystem content domain on the X platform.

  • Refine the AI Agent curation evaluation system, complete the integration with Platform X, and build a content retrieval engine for Platform X.
  • Based on the X platform, fully implement the core process mechanism for AI Agent curation and tipping.

Risk Points:

  • Joyid’s support for the Fiber Network is currently under development and testing. If it is not completed during this period, the reward method will temporarily switch to using the CKB chain.
  • Users who receive tips are required to provide their receiving addresses, which can make the early experience cumbersome and potentially affect the tip conversion rate. We will focus on optimizing the user address binding service on the X platform.

Cold Start Strategy

  • Moderately lower the evaluation standards for agents to enhance the early user experience.
  • Hunter’s Reward: Users who recommend high-quality, undiscovered content will receive a 10% cut if the AI tips the content later.
  • Curated Gamification: Release “curated tasks” (such as “Best Intro to Quantitative Trading”) weekly, and reward the winner of each task.

2. Beta Phase (2-3 Months):

Expand the scope of curated and rewarded content, enhance influence and user scale, and improve product features and user experience.

  • Enhance the identification mechanisms to prevent witch attacks and malicious tipping, ensuring the healthy operation of Agents.
  • Implement extended features such as user-initiated tips and curation, and promote them to the user base of the X platform to create a benchmark effect.
  • Accessing the Fiber Lightning Network for tipping, optimizing the tipping process and user experience
  • Collaborate with project teams and communities to curate content in specific fields, expand the user base, and grow the donation pool.
  • Targeted tipping and curation for Web3 KOLs to amplify project influence.

3. Formal Operation Stage (Long-term):

Comprehensive coverage of high-quality content rewards and curation in the Web3 and Crypto fields, centered around users’ real needs, transforming it into a practical AI Agent application that seamlessly integrates into everyday scenarios.

  • Continuously promote and expand influence and user base, making “tipping” and “curation” deeply ingrained in users’ minds - these are the key focuses for the future.
  • By attracting sponsors to the tipping pool through content exposure, we encourage sponsorships.
  • Expanding Agent services to platforms like Discord and Telegram to broaden the user base and enhance influence.
  • Encapsulate the tipping capability for other users and Agent applications to invoke, establishing an Agent ecosystem position in the tipping domain.
  • Staying ahead of AI and AI Agent technology advancements, integrating innovative Web3 community practices, and continuously driving progress.

4. Budget Application and Allocation

This proposal plans to request a total of 81,000U from the CKB Community Fund DAO, which translates to 18,629,254 CKB based on the CKB price of 0.004348 at the time of proposal preparation.

1. Budget Details

Category Budget Description
Labor Subsidy Cost 36000U Estimated 1 full-time developer + 1 full-time operator, 1500U per month per person x 12 months, part-time at 50%.
Development Cost 20000U Including regular back-end service deployment, X platform development account, large model token consumption, and third-party services required for development, service, and operation.
Operation Costs 25,000U Of which 10000U are used for the Agent Reward Pool, and the remaining is used for the cold launch period, event promotion period and long-term operation and promotion.

All budget categories will be allocated and utilized progressively over a period of one year or longer, as detailed below:

  1. Personnel Subsidy Costs (36,000 U):
  • Monthly subsidy of 1500U per person, covering 1 full-time developer and 1 full-time operator, disbursed linearly on a monthly basis (12-month subsidy period).
  • If team members are part-time, the subsidy is halved (750U/month).
  1. Development Costs (20,000 U):
  • 【20%】Backend application deployment server, database, etc. expenses
  • 【45%】X Platform development account fee【Due to the high cost of official services on the X Platform and its extended content engine services】
  • 【15%】AI large model API call fee
  • 【10%】Third-party service subscription fees (e.g., monitoring, data analysis, etc.)
  1. Operating Costs (25,000U):
  • 【40%】Agent Reward Pool Funds: 10000U 【The maximum amount allocated from the budget to the reward pool, aiming to achieve open-source income growth before depletion】
  • 15,000 units are allocated for:
    • 【6%】Cold Start Promotion Campaign
    • 【9%】Active User Incentives
    • 【45%】Expenses for long-term content marketing, user incentive campaigns, and community operations lasting a year or more.

The team commits to documenting and disclosing all financial transactions in detail, ensuring transparency and effectiveness in the use of funds.

2. Milestones and Progress Assessment Mechanism

To ensure steady project progress and efficient use of funds, we have established the following phased milestone mechanism and corresponding evaluation indicators:

Phase Conditions of Release Specific Amount Estimated Time Period CKB Conversion
Milestone0 Issued after the proposal is approved for project initiation 10000U After proposal approval 2299908
Milestone1 Completion of basic framework construction, realization of core functions, official launch on the X platform and implementation of tipping and curation capabilities. 13000U 1-2 months 2989880
Milestone2 Completion of online user-initiated tipping, active curation expansion and other functionalities, and accumulation of 500+ concerned users on X platform. 13000U 2-3 months 2989880
Milestone3 Accumulate 1.8K+ concerned users of X platform, real users use active tipping ability 10 times+, real curated reward 500+ contents. 14000U 1~2 months 3219871
Milestone4 Accumulate 5K+ followers of X platform, 1K+ real content rewarded, real users using active tipping ability 100 times+, accept 3+ Sponsor sponsorships. 15000U Around 2 months 3449862
Milestone5 Accumulate 10K+ followers on X platform, increase reward pool by 3000U+. 16000U 2~3 months 3679853

Launch Phase (Milestone 0)

Initial funding of 10,000U will be allocated for setting up the basic development environment and forming the team. This fund will be disbursed immediately upon proposal approval to expedite project initiation and drive the development and implementation of core features.

Foundation Framework Phase (Milestone 1)

Complete the development of core functionalities and achieve the first deployment on the X platform. Specific evaluation metrics:

  • Implement an AI Agent-powered content curation and evaluation system, complete the integration with the X platform, and build a content retrieval engine for the X platform.
  • Successful implementation of the curation and reward process for AI Agents on the X platform.
  • 13000U will be issued upon completion.

Feature Extension Phase (Milestone 2)

Enhance product features and validate market acceptance. Evaluation metrics:

  • Develop and deploy features for user-initiated tipping and curation.
  • Implement security mechanisms to prevent witch attacks and optimize the user experience for the tip process.
  • Integrate and successfully deploy the Lightning Network tipping process based on Fiber.
  • Achieve 500+ authentic user followers and complete 80+ content rewards.
  • 13000U will be issued upon completion.

Additional Notes:

We understand the importance of setting milestones and clear assessment criteria for community review proposals and oversight of fund usage. Unlike previous application-oriented proposals, past approved proposals primarily focused on the development and implementation of applications with clear functionalities, allowing for simple milestone setting based on functional modules for assessment, with less involvement in post-launch usage. However, our Agent application has a longer cycle, with plans to use the budget in stages over a year to advance systematically. Although the initial milestones 1 and 2 emphasize technical implementation and delivery, in the long run, continuous operation and iterative capabilities are crucial, with market recognition and user popularity serving as our ultimate evaluation metrics.

The subsequent three milestones will initially be evaluated primarily through quantifiable operational data indicators. While we will certainly continue to iterate on our technology and expand our platform, the specific timing of implementation will depend on product and market conditions, and thus has not been included in the current evaluation metrics. Although the operational metrics we have set are not overly aggressive, they ensure steady progress in the project. If the community has more reasonable and constructive suggestions regarding the milestones and evaluation metrics, we warmly welcome them to be proposed during the proposal discussion phase.

Market Validation Phase (Milestone 3)

Validate product-market fit. KPIs:

  • Accumulated 1800+ real users’ attention.
  • Users actively tip more than 10 times.
  • The AI curation system has completed over 500 effective content rewards.
  • Issued 14,000U upon completion.

Scale-up Phase (Milestone 4)

Achieve breakthroughs in user scale and commercialization. Key performance indicators:

  • Over 5000+ real users following
  • Completed over 1000 content tips
  • Receive sponsorship from at least 3 individuals/organizations/project parties.
  • Issued upon completion: 15,000 U.

Ecological Construction Phase (Milestone 5)

Establish a sustainable development model. Assessment indicators:

  • Over 10K users accumulated
  • The tipping pool has achieved autonomous growth, exceeding 3000U.
  • 16000U to be issued upon completion.

Phase Transition and Risk Management

Subsequent operational outcomes are subject to market conditions and the popularity of Agent application development, carrying inherent uncertainties. In the event that targeted milestones are not met on schedule, we will:
• Strengthen community feedback mechanisms and appropriately extend milestone deadlines.
• Adjust user growth trajectory strategies to ensure the ultimate objectives remain unchanged.

If market validation indicates that the current stage of Agent applications fails to meet effective demand, falling short of the anticipated scale and impact, we will terminate subsequent budget requests and return the remaining funds.

5. Other Instructions

Who am I, and why am I the one to do this?

I am an independent developer who has worked at internet companies such as 360 and Ant Group. In the previous market cycle, I participated in developing an NFT rental protocol and was responsible for the issuance and operation of an NFT project. I am well-versed in blockchain technology and the Web3 ecosystem, firmly believing in the long-term value of blockchain and continuously focusing on the Web3/Crypto domain.

The inspiration for this proposal came from an idea I had after signing up for the Rock Web5 Hackathon, aiming to integrate AI Agents with a tipping mechanism. During the hackathon, I completed a minimal viable product (MVP) for curation and tipping based on the Nostr Protocol. After in-depth discussions with my peers and receiving valuable assistance and feedback from the project team, I went through multiple rounds of refinement, resulting in the current version.

If this proposal is approved by the CKB community, I commit to developing under my real name on social media, regularly disclosing project progress and budget usage, and fully driving the project’s development. If the proposal is not passed, I fully respect the community’s decision. CKB community members are welcome to provide valuable feedback on all aspects of this proposal!

5 Likes

Hi,
Isn’t Silent Berry building something similar, how are you different from them? Who will the creators be, where will they come from? Can you guarantee them or will it just be another project, ready at the protocol level, waiting for its users to come?

Hi D-lyw, extremely well put together proposal and I think a very good idea.

As I was reading I was wondering mainly about one thing, which you have also thought about:

To me, this will be the hardest and most important part because you can’t rely on people opting in to use CKB, you’ll need to somehow force it on them and make it worthwhile for them to claim their tips.

I almost hate to say it, but a protocol token might also be required to bootstrap users into the system, with claiming tips (and also tipping) being ways of securing a portion of the future airdrop.

Anyway, I’ll vote for this for sure, good luck!

Glad to meet your comments
Our Agent app and Silent Berry should be completely different positioning and type of products. Silent Berry is a Web3 publishing and distribution platform that focuses on on-chain distribution of books, leveraging the decentralized and immutable nature of blockchain.
And our Agent application, based on the AI Agent feature of AI and the payment ability of the Fiber Lightning Network, do the curation and tipping. You can simply understand that it is a bot of the X platform, which can actively tip high-quality content in various fields of the X platform, and its evaluation criteria for the content are completely fair and transparent. For a similar Agent application form, please refer to AIXBT (an AI Agent application in the web3 field focusing on market analysis and perception)

1 Like

Glad to see your message, in a way, you’re right.
However, I am worried that the introduction of Token too early will most likely accelerate the time of frenzy and death, and shorten the life cycle of the entire project. In the early stages of the project, we will focus more on the implementation of the system and optimize our entire system, and there is no similarly positioned Agent application in the current market, which gives us some opportunity to gradually promote our application in the cold start stage to create real long-term use value. However, as a player in the Web3 field, I personally respect the Token play of Web3, and if there is a chance, it is one of the potential ways to operate and promote it in the long run

I 100% agree, launching a token for no other reason than just because this is crypto is a bad idea for the longterm success of the project.

But you also need to understand that nearly all CKB projects have a proven track record of a short life cycle, so unfortunately, all the data points to a very slim chance of longterm success and adoption anyway.

The truth is you just can’t rely on the existing CKB community and the Nervos core team to help you.

The already small community is mostly made up of holders, rather than blockchain users and the Nervos team offers no ongoing promotional support to ecosystem projects, so the trick is to break out of this box and into the broader crypto community, which is extremely difficult.

On a positive note though, I think your project actually has a real chance at doing this, but you are going to need to create your own hype and like it or not, a token is probably the best way to do this.

But it needs a real use case that’s implemented from day one of the token launch, one obvious method would be to tie the token to the claiming of the tips, for example:

  • To claim any tips, the user needs to submit 2% (or whatever percentage you decide) of the total tip value in the protocol token.
  • The 2% of protocol token is permanently burnt and removed from circulation.

Another option would be that when a user pays a tip in the protocol token, 2% of the amount is burnt.

If you could also create a way that the protocol token was used as the primary tip method, but it was swapped on the backend to USDI/USDT, that would be even better.

There’s probably lots of other good ways to implement some real use cases, just needs a bit of vision, so I don’t think you should discount it.

2 Likes

Hi @D-lyw

Thank you for the proposal

I like the idea generally, streamlining and automating tasks using AI opens up lots of creative options. I think the community would really value a tipping bot to reward contributors, and I think AI-guided content curation also has merits. I have some thoughts and follow up questions:

  • What LLM model will you be using to build this agent?

  • Aixbt, which you cited for reference, on cursory glance seems to assess on chain metrics, market dynamics, and social media activity. Could you share more in-depth information around the multimodal assessment algorithm (Information Entropy, Originality, Community Value), how these metrics are assessed, and your research regarding its strengths and weaknesses?

    • I see how this could be generalised to the wider industry, as you mention in Stage 3 of your plan (Formal Operation Stage), although that may require additional funding streams or grants from other projects. What is your GTM strategy to appeal to the wider market and promote your product?
  • Is this agent and tipping bot limited to X only? Some of these features would also be useful in community groups that live on Telegram such as Nervos Nation.

  • Have you researched other tipping bots and their user flow? I have seen some in other utxo communities, it may be worth researching how they address UX challenges.

  • When it comes to milestones, in my view payments must be linked to development-based milestones and not growth-related. This is because unfortunately it’s easy to generate activity and numbers of X followers through inorganic means. Perhaps you may opt to do it to kickstart momentum (I see arguments for and against), but I think these kinds of metrics should be separate from any grant. In that respect, I would like to see those milestones adjusted to reflect development or operations.

  • Will this project be open-sourced? I think it would be beneficial to the community as there are interesting lego pieces here.

  • Lastly, I agree with you that there isn’t really a need for a new token here.

It’s all well and good to take the anti-token highroad here, but please just explain to me where the excitement and the users will come from.

And how people from outside the CKB ecosystem will be encouraged to create a Joyid wallet and claim their 20 cent tip.

There’s absolutely nothing wrong with tokens that have a real use case and have a mechanism to stop them being inflated to hell.

At this stage of crypto, they are almost essential to get an independent project off the ground and give it a chance for adoption.

And unlike most projects, there’s actually a crystal clear usecase here to create a tipping token.

Even if the price almost goes to zero and the token is considered worthless, then people will actually use it more and be giving out very generous tips, which means more burn, less supply and then the next thing you know the price is going up again.

Hi neon, it’s a pleasure to see your comment.

During the hackathon, I utilized OpenAI’s LLM. This was because our early Agent’s content review process was primarily achieved through Prompt Engineering. The performance of mainstream LLMs should be relatively similar, with the primary consideration being cost, as it constitutes a long-term, ongoing expense. Moving forward, we will consider adopting models like Deepseek, which meet our criteria while offering a more cost-effective solution.

In the early stages, we used LLM’s prompt word engineering to review and examine dimensions such as information entropy, originality, and community value. When we have the opportunity to continue to develop, we will consider fine-tuning the training of a small model to complete the review of content, or introduce ZKP to validate the evaluation process of the AI to avoid cheating or manipulation. The core competitiveness of using AI Agent for content review lies in the openness, transparency and verifiability of evaluation criteria, unlike the centralized black box operation of Web2 social media

Of course not, we need to complete our core mechanism on the X platform first to verify the fit between market feedback and user needs. Once the market proves that this is a real and effective agent application, we will promote more platforms such as Discord and Telegram to expand the user base and influence

Yes, a good user interaction experience is an important part of the success of the product. The current tentative process is that when the user is tipped by the Agent for the first time, the user needs @Agent and leave the address where the user receives the tip. The subsequent reward will be directly tipped to the address, and appropriate comments and reminders will be left.

As mentioned in the proposal, we expect to be able to complete the development of all system modules set up prior to the official operational period in a timeframe of no more than 5 months (excluding expansion to other platforms). But there’s no point in just developing it and no one using it. Our Agent application is not only for the CKB community, but also for the entire Web3/Crypto ecosystem users, which is a very challenging goal, in fact, whether our Agent application is recognized by the market, whether the effect is good or not, the community users are easier to perceive and judge. For the development tasks in the subsequent stage, from a responsible point of view, it is not possible to write the deterministic development tasks for the subsequent six months in the one-year plan in the proposal at this stage, and it needs to be adjusted according to the actual progress and feedback from market users.
Of course, I also very much agree with the problem you pointed out: the number of followers and users of the X platform is easy to fake. In addition, after completing the Alpha and Beta phases, we can re-communicate with the community and align the corresponding tasks, milestones and assessment standards for the next six months based on the actual progress and market effect.

Of course, once we have completed our core systems and frameworks, we welcome more users to participate. In particular, regarding the content evaluation criteria of the Agent model, the model bias is corrected and adjusted based on community voting. Users are also welcome to try more extended gameplay based on the capabilities of our Agent

2 Likes

Sorry Yeti, I didn’t mean it as a dig if that’s how it came across :laughing:

It’s mostly my perspective that currently the market sentiment doesn’t seem to be right for airdrops and is unlikely to be for the time being. I do wish it was a better climate. Airdrops can work for projects with an existing, established market presence (and even then interest is usually short-lived). Less so for new projects.

Hey mate, all good, I totally get your point and I agree for the most part, I’m not a fan of tokens either when there’s no good reason, it never ends well other than for the founders and the early buyers/insiders.

I just think this particular project has potential for a really good implementation that can be managed well.

If you can combine a good usecase with good tokenomics, it’s a great way to generate a big community quickly, then you only need a good product to keep the community active and interested.

1 Like