筛选结果
11 篇排序
51:31ENThey Refunded Nearly $2 Million. Then Factory’s CEO Explained the Dark FactoryMatan Grinberg on being two years early, why a model-agnostic harness beats lab co-design, and why 90 percent of coding tokens will soon run with the lights off.
51:09ENBuild like the model will eat your harness in a yearLogan Kilpatrick, who leads Google AI Studio and the Gemini API, tells Sequoia's Sonya Huang why agent scaffolding is a 12-month asset, why coding already feels like narrow superintelligence, and why Google is optimizing for outcomes instead of eyeballs.
45:33ENCursor Ships 1TB of Composer Weights Across Oceans in Under a MinuteFederico Cassano and Fireworks' Dima walk through the globally split RL stack that trained Composer 2 on a sparse trillion-parameter MoE.
48:03EN → 简体中文编程终将被解决,真正的瓶颈是说清楚你想要什么Warp 创始人 Zach Lloyd 向红杉资本的 Sonya Huang 讲述,为什么重做后的终端恰好成了编程智能体的合适形态,以及他为什么不再为 Token 用量提供补贴。
40:13ENAs long as humans still need to read code, the IDE is not dyingZed founder Nathan Sobo argues that source code is a language for people, that terminal agents still need a visual editor, and that conversations should be anchored to every edit.
1:01:00ENv0 Hit 3 Million Users. Vercel’s CEO Says Pages Will Be Generated On Demand Within Five YearsFor a company that sells cloud hosting and frontend tools, one of the fastest-growing signup channels is ChatGPT.
37:45ENCodex Is a Thought Experiment: What If the Agent Worked on Its Own Computer?OpenAI’s Codex team on training o3 to write mergeable PRs, running four copies of a 1 a.m. bug, and why the product only clicks once you stop pairing and start delegating.
1:26:19EN → 简体中文为了一个有十亿开发者的世界,他拒绝了 10 亿美元Replit CEO Amjad Masad 在红杉资本的《Training Data》节目中谈到:一家只有六个人的公司,为什么放弃了足以改变人生的收购报价;以及在他看来,十亿人都来写软件,比把 AGI 当作全民基本收入(universal basic income),是更好的终局。
51:15EN → 简体中文通用人工智能只剩一个难题:在单个 Token 上多花一百万倍的算力Magic 创始人 Eric Steinberger 在红杉资本 Training Data 节目中说明:他为什么要自己训练模型,为什么软件工程师是通用人工智能(AGI)的 MVP,以及为什么 96% 仍然算不上同事。
1:07:34EN → 简体中文GitHub 起初不信,Copilot 已写下 25% 的代码GitHub 首席执行官 Thomas Dohmke 向红杉资本讲述:一个计划外的 GPT-3 实验如何变成 Copilot,GitHub 为何至今仍不自己训练基础模型(foundation model),以及他希望到 2030 年让十亿人写软件。
59:11EN → 简体中文别训练模型,Factory 创始人说应该去淘金Matan Grinberg 和 Eno Reyes 在红杉资本的《Training Data》节目中谈到,他们为什么要在别人的模型之上,卖一整套专门负责具体工作的 droid;这套技术栈刚刚把 SWE-bench 的成绩从 14% 提到了 19%。









