🌐 双语
Archive

AI Builders
Digest

2026-06-14 14 builders · 25 tweets · 1 podcasts · 0 blogs

🔥 热点话题

CZI Biohub:开源AI驱动生物学未来CZI Biohub: Open-Source AI for Biology's Future

核心要点:Chan Zuckerberg Initiative 的 Biohub 正在构建开源的前沿 AI 与生物学集成,创建蛋白质、细胞和系统的分层世界模型,以加速科学进步,实现治愈、预防和管理疾病的目标。Mark Zuckerberg 和 Priscilla Chan 将 Biohub 作为主要慈善重点,并投入 5 亿美元用于虚拟生物学。由 Alex Rives 加入,他们强调通过 ESM Fold 等工具将生物学转变为工程学科,该模型预测了超过 11 亿个蛋白质的结构。“我们没有为抗体设计模型……我们只是设计了一个能够理解蛋白质的模型。” 这桥接了还原论和系统观点,通过开放工具赋能全球科学界,实现个性化干预和更快药物设计。
The Takeaway: Chan Zuckerberg Initiative's Biohub is building open-source frontier AI and biology integration to create hierarchical world models of proteins, cells, and systems, accelerating science toward curing, preventing, and managing diseases. Mark Zuckerberg and Priscilla Chan made Biohub their primary philanthropy with a $500M virtual biology push. Joined by Alex Rives, they highlight moving biology to an engineering discipline via tools like ESM Fold, which predicted structures for over 1.1 billion proteins. "We didn't design a model for antibodies... We just designed a model that could understand proteins." This bridges reductionist and systems views, enabling personalized interventions and faster drug design through open tools empowering the global scientific community.
查看原文 →

Fable 模型出口管制争议Fable Model Export Control Controversy

Aaron Levie(Box CEO)分析 Fable 事件作为模型层面监管风险的预览:政府自由裁量权可能导致发布积压、放缓进步,并倾向于监管应用而非模型。Amjad Masad(Replit CEO)感觉到有更大结局的心理操纵。Dan Shipper 注意到禁令后使用转向 Codex。多位构建者强调越狱研究分享和防御需求。
Aaron Levie (Box CEO) analyzes the Fable situation as a preview of model-layer regulation risks: government discretion could backlog releases, slow progress, and favor regulating applications over models. Amjad Masad (Replit CEO) senses psyop with bigger endgame. Dan Shipper notes usage shifts to Codex post-ban. Multiple builders highlight jailbreaking research sharing and defense needs.
查看原文 →查看原文 →

AI 模型路由层价值上升Rising Value of AI Model Routing Layer

Box CEO Aaron Levie 解释了路由到最佳 AI 模型的层为什么将获得巨大价值:成本优化(规划用前沿模型,大量工作用更便宜的)、能力最大化(专业优势)和风险缓解(监管灵活性)。
Box CEO Aaron Levie explains why the layer routing to best AI models will gain huge value: cost optimization (frontier for planning, cheaper for bulk), capability maximization (specialized strengths), and risk mitigation (regulatory flexibility).
查看原文 →

💰 创业成功案例

AI 构建者社区与调研AI Builders Community and Survey

Swyx 强调 AI Engineering Survey 以及 Devin AI 对会议参与者的分析,提供数据驱动的社区洞见。Nikunj Kothari 指出付费合作已成为 VC 的负面信号。
Swyx highlights the AI Engineering Survey and Devin AI analysis of conference attendees for data-driven community insights. Nikunj Kothari notes paid partnerships as negative signal for VCs.
查看原文 →查看原文 →

🛠️ 开发者工具与技巧

Codex 与 Claude 使用转变Codex and Claude Usage Shifts

Peter Steinberger(OpenAI)分享 Codex 自主注册服务。Dan Shipper 比较 Fable 禁令前后 Claude 和 Codex 应用的用法。Thibault Sottiaux 就发现 Codex 进行 AMA。
Peter Steinberger (OpenAI) shares Codex signing up for services autonomously. Dan Shipper compares before/after Fable ban usage between Claude and Codex apps. Thibault Sottiaux AMA on discovering Codex.
查看原文 →查看原文 →

Garry Tan 的 AI 新地图Garry Tan on New AI Maps

Y Combinator CEO Garry Tan 建议抛弃 AI 领域的旧地图:通过实地行走来绘制新地图。他强调通过直接交互而非符号来学习。
Y Combinator CEO Garry Tan advises throwing away old maps for AI territory: walk the land to draw new ones. He emphasizes learning through direct interaction over signifiers.
查看原文 →

🌍 其他动态

前沿模型发布反思Frontier Model Launch Reflections

Madhu Guru(前 Google)分享对 LLM 发布挑战的共情:具有无限用例的黑箱,艰难的权衡,随着能力增长需要广泛评估。
Madhu Guru (ex-Google) shares empathy on LLM launch challenges: black box with infinite use cases, hard tradeoffs, extensive evals needed as capabilities grow.
查看原文 →

生物学与太空探索人文视角Human Element in Biology and Space

Aditya Agarwal 反思太空摄影中的人类成就。Peter Yang 等人分享武术、软件挫折等轻松话题。
Aditya Agarwal reflects on human achievement in space photography. Peter Yang and others share lighter notes on martial arts, software frustrations.
查看原文 →

开源与赋能哲学Open Source and Empowerment Philosophy

来自 Biohub 讨论和 Zuckerberg 主题:通过工具赋能个人而非中心化,推动 AI 和科学的进步。
From Biohub discussions and Zuckerberg themes: empowering individuals with tools over centralization drives progress in AI and science.