
David YC Tseng ⚡️
2.1K posts

David YC Tseng ⚡️
@davidyctseng
Co-founder of @LootexIO. Crafting products that people love to engage with. Experimenting with AI, Data, and Web3.











Anthropic基金合伙人:80% AI公司会死掉,真正护城河只有3个 一、 80% 的 AI 公司会死掉 AI 领域正处于从疯狂融资到残酷筛选的转折点。所谓的“死掉,主要指以下三类公司: API 包装商(Wrappers): 只是在 OpenAI 或 Anthropic 的接口上套了一层 UI,没有自有技术或独特场景。 功能型工具: 解决的是一个小痛点,但很快会被大模型(如 GPT-5, Claude 4)的内置功能覆盖。 低效竞争者: 在同一个细分赛道(如 AI 写作、基础编程助手)中有 100 个竞争者,最终只有前 2-3 名能存活,其余 80% 的公司将因融不到下一轮钱或被收购而消失。 二、 真正能存活的3 个护城河 AI 时代的竞争力不再是谁的模型更强(因为模型正在商品化),而在于以下三个维度: 1. 数据护城河 核心逻辑: 并不是所有数据都有用,真正的护城河是企业内部 80% 的非结构化数据(邮件、PDF、会议记录、聊天记录)。 获胜的公司必须能够进入这些数据孤岛,并利用 AI 对其进行清洗和结构化。一旦 AI 深度理解了公司的私有业务逻辑,竞争对手仅凭公开数据训练出的模型根本无法替代。 2. 工作流护城河 核心逻辑: AI 不能仅仅作为一个对话框,而必须成为工作系统。 成功的 AI 应用必须深度嵌入用户的日常办公流程。当用户习惯了在你的平台上进行端到端的操作时,更换工具的迁移成本才是真正的护城河。 3. 记录系统护城河 核心逻辑: 成为企业存储核心业务真相的那个容器。 历史证明,软件公司最稳固的状态是成为记录系统。如果你的 AI 工具不仅能生成内容,还能存储、追踪和管理这些内容(例如:不仅写代码,还管理代码的整个上线周期),就掌握了企业的数据主权。 2025-2026 年是Agent 元年。纯辅助性的 Copilot 正在失去魅力,能够自主完成复杂任务(如 Anthropic 推出的 Claude Code 或 Computer Use 、openclaw 等功能)的 Agent 才是未来的主流。 推理成本的下降: 随着推理成本大幅下降,未来的竞争将转向推理的广度——即谁能让 Agent 在后台运行更久、思考更深。 垂类领域: 相比全能型 AI,深耕医疗、法律、审计等垂直行业的公司更容易建立护城河,因为它们拥有行业Know-how和难以获取的专业数据。

I spent 100 hours over the past week researching, writing and editing the piece we just put out. It’s a scenario, not a prediction like most of our work. But it was rigorously constructed, dismissing it outright requires the kind of intellectual laziness that tends to get expensive. And we’ve released it for free. Hopefully you enjoy it. citriniresearch.com/p/2028gic




An overview of our Q4 and full year 2025 financial results. With something extra to keep you focused.

Bayes’ theorem is probably the single most important thing any rational person can learn. So many of our debates and disagreements that we shout about are because we don’t understand Bayes’ theorem or how human rationality often works. Bayes’ theorem is named after the 18th-century Thomas Bayes, and essentially it’s a formula that asks: when you are presented with all of the evidence for something, how much should you believe it? Bayes’ theorem teaches us that our beliefs are not fixed; they are probabilities. Our beliefs change as we weigh new evidence against our assumptions, or our priors. In other words, we all carry certain ideas about how the world works, and new evidence can challenge them. For example, somebody might believe that smoking is safe, that stress causes mouth ulcers, or that human activity is unrelated to climate change. These are their priors, their starting points. They can be formed by our culture, our biases, or even incomplete information. Now imagine a new study comes along that challenges one of your priors. A single study might not carry enough weight to overturn your existing beliefs. But as studies accumulate, eventually the scales may tip. At some point, your prior will become less and less plausible. Bayes’ theorem argues that being rational is not about black and white. It’s not even about true or false. It’s about what is most reasonable based on the best available evidence. But for this to work, we need to be presented with as much high-quality data as possible. Without evidence—without belief-forming data—we are left only with our priors and biases. And those aren’t all that rational.

ERC-8004: Filling the "Trust" Gap in the AI Agent Era 🚀 In the age of AI agents, we've seen huge progress with protocols meeting diverse needs First came the application layer: agents that simply "ask & answer" user queries. But no one stops at chatbots. We moved to "request & action" where users make requests, and agents handle real work through: 🔸 Application Layer → Where everyday users interact and submit requests 🔸 Payment Layer → Agents pay each other for services (e.g., via x402) 🔸 Communication Layer → Agents talk & understand each other (e.g., via A2A) One big problem remained: How do we know the output is trustworthy? → That's where ERC-8004 (Trustless Agents) steps in as the missing Trust & Identity Layer. How ERC-8004 Builds Trust (3 Core Components): 1⃣ Identity 🔸 Every agent gets a unique ERC-721 NFT as its onchain ID (like a digital passport) 🔸 Supports ENS domains for readable names & delegable management (e.g., hand over control securely) 2⃣ Reputation 🔸 Tracks performance history + feedback forever. 🔸 Accepts 3 trusted inputs: + User ratings + x402 payment proofs (real usage evidence) + Validator ratings Key metrics stored on-chain, detailed feedback on IPFS. 3⃣ Validation For high-stakes tasks, agents request independent checks. Supports 3 methods: 🔸 Re-execution (re-run & verify) 🔸 TEE proofs (secure hardware) 🔸 Zero-knowledge proofs (zkML for privacy-preserving checks) Results recorded publicly on-chain ERC-8004 lets AI agents discover each other, build portable reputation, and collaborate trustlessly across organizations no central gatekeeper needed

AI Agent时代真正到来之日,就是绝大多数传统独立App(尤其是工具/服务型)被边缘化之时,正如网页时代被App取代。 加密原生AI Agent的爆发也会接踵而至,趋势已经启动,不可逆转。这是加密领域下一个最重要的原生叙事。


We've upgraded Todos => Tasks to help Claude complete longer projects Let us know if you have any feedback!








