筛选结果
10 篇排序
1:04:14EN → 简体中文帮忙在 120 天内建起 Colossus 的人,正押注一个住在你家里的 AI主持 Jacob EffronxAI 联合创始人 Igor Babuschkin 离开后创办了 River AI。他认为,编程智能体(Agent)已经跨过了一道门槛,闭源实验室正被挤压,而下一个大奖是个人化、本地化的智能。
1:13:33EN → 简体中文一万亿个 Token 之后,概念终于落地:Łukasz Kaiser 谈 Transformer 仍然缺什么主持 Jacob Effron《Attention Is All You Need》合著者在 Unsupervised Learning 节目中表示,Transformer 大概率会继续胜出,但它们学习的方式仍然与人不同。
57:37EN → 简体中文OpenAI 企业产品负责人:换模型只改一个参数的日子已经过去主持 Jacob EffronOlivier Godement 谈 GPT-5.1、Codex、企业用的骨架层(harness,即把模型接到工具上的那一层),以及真正让 AI 在大公司落地的三条做法。
42:06EN → 简体中文SWE-bench 突破 80.9% 之后,Dianne Na Penn 如何评价 Opus 4.5主持 Jacob EffronAnthropic 负责计算机使用研究的产品负责人,谈有误导性的 token 价格、轻量框架(harness),以及为什么对齐(alignment)会让模型成为更好的思考者。
59:06EN → 简体中文推理成本降了 100 倍,Tri Dao 说还能再降 10 倍,英伟达 90% 的份额守不住主持 Jacob EffronFlashAttention 作者、Together AI 首席科学家谈为什么硬件可移植性是个神话,Claude Code 如何把他的效率提升到原来的 1.5 倍,以及为什么工作负载会瓜分芯片市场。
57:31EN → 简体中文衡量模型的新标尺:它们能独立工作多久主持 Jacob EffronClaude 4 发布当天,Anthropic 研究员 Sholto Douglas 认为,AI 的下一阶段不再主要比谁能解更难的谜题,而是比谁能在无人监督的情况下持续工作数小时。
1:11:06ENReplit's Amjad Masad on Building Your Own Model, Latency, and the Next Billion Software Creators主持 Jacob EffronThe Replit founder explains why AI should be native to the product, why he trained a small in-house model, and why he thinks software engineering is about to split in two.
1:04:25ENInside GitHub's Copilot Bet: Agents, Open Source, and Why the Developer Job Endures主持 Jacob EffronGitHub CEO Thomas explains how Copilot went from a research bet to 1.3 million developers, and what agents, fine-tuning and a multi-model world mean for the next five years of software.
1:36:57ENDo the Dumb Thing First: A CTO's Playbook for Building an AI Coding Assistant主持 Jacob EffronSourcegraph CTO Beyang Liu on which AI coding features actually earn daily use, how new models get shipped and evaluated, and why retrieval is still a search problem.
54:52ENCoding Agents Are Breaking Containment: The 2026 AI Stack, the Coding Wars, and the Next Frontier主持 Jacob EffronJacob Effron and Swyx map the state of AI engineering, from agent harnesses and infrastructure churn to coding-market economics, open models, dark factories, and world models.









