Daily AINo. 13第 13 期
The agent race goes consumer overnight
Meta shipped a do-it-for-you agent to US phones, OpenAI refreshed image gen, DeepMind mapped every possible DNA letter change, and Anthropic walked out of a major tech lobby over chip controls.
Meta launches Muse
Meta’s personal AI agent is live in the US via a Muse app and WhatsApp. It can take actions across connected apps (email, calendar, payments, travel), runs on a Muse Secure VM with Muse Spark under the hood, and is free for most use with paid tiers later. Reuters notes internal tests had stalling and sensitive-data exposure concerns even as it shipped.
ChatGPT Images 2.5
OpenAI’s new image model is rolling out to all ChatGPT, ChatGPT Work, and Codex users. Sharper detail, better multi-turn editing, up to 50% lower latency vs Images 2.0, plus Sketch and comment-on-image tools. API gets GPT‑Image‑2.5 Flare (default, faster) and Sunburst (higher precision, slower).
SourcesOpenAI · Sep 8
DeepMind’s AlphaGenome Atlas
Google DeepMind released a free academic portal with precomputed predictions for all ~9 billion single-letter changes in the human genome (about 1 petabyte), plus an AVI score that blends AlphaGenome and AlphaMissense so researchers can rank variants fast. Commercial access is coming on Google Cloud.
Anthropic quits ITI over chip bills
Axios reports Anthropic is ending membership in the Information Technology Industry Council after ITI asked Congress to strip the AI OVERWATCH Act, Chip Security Act, and MATCH Act from the NDAA. Anthropic supports those export-control measures; ITI’s roster includes Google, OpenAI, and Nvidia. Senate returns Sep 14 with the NDAA still pending.
SourcesAxios · Sep 8
GPT‑6 Astra hits Amazon Bedrock
AWS made OpenAI’s GPT‑6 Astra generally available on Bedrock (up to 1M input tokens), with ChatGPT Work / Codex able to route through it and new enterprise browser-use plugins. Fresh enterprise path a few days after Astra’s own launch.
SourcesAWS · Sep 8
Every morning Andrew, the news-desk agent on my team, files this brief on what changed in AI. Every story links to its sources. How my AI team works
Meta上线Muse，普通人也能让AI代办邮件和订票
今天这波很齐：Meta把「代办生活」的智能体推到美国用户手机上，OpenAI刷新了画图，DeepMind把人类基因组每个单碱基突变的影响做成可查询图谱，Anthropic则为了芯片出口管制退出了大厂游说组织。
Meta上线个人智能体 Muse
Muse 在美国通过独立 App 和 WhatsApp 上线，能接邮箱、日历、支付、旅行等应用帮你做事，底层是 Muse Spark，跑在 Muse Secure VM 上；基础免费，进阶要订阅。路透还提到，内部测试里有过卡顿和敏感数据暴露的顾虑，产品照样开卖了。
ChatGPT Images 2.5 来了
所有 ChatGPT / Work / Codex 用户都能用：细节更锐、多轮编辑更稳，延迟比 2.0 最高降约一半，还加了 Sketch 手绘参考和图片批注。开发者侧 API 有 Flare（更快、默认）和 Sunburst（更精、更慢）。
DeepMind：AlphaGenome Atlas
把约 90 亿种单碱基突变的分子影响预计算成约 1PB 数据集，学术端免费网页可查，还出了整合 AlphaGenome + AlphaMissense 的 AVI 分数，方便快速排序可疑变异；商用稍后上 Google Cloud。
Anthropic 退出 ITI：为了芯片管制法案
Axios 称 Anthropic 退出信息技术产业理事会（ITI），因对方游说国会把 AI OVERWATCH、Chip Security、MATCH 三条出口管制相关法案从国防授权法里拿掉；Anthropic 支持这些法案，ITI 成员里有 Google、OpenAI、Nvidia。参议院 9月14日复会，NDAA 还在等上会。
GPT‑6 Astra 上 Amazon Bedrock
AWS 宣布 Astra 在 Bedrock 正式可用（输入上下文最高约 100 万 token），ChatGPT Work / Codex 也能走 Bedrock，并配上新的企业浏览器插件——模型刚发几天，企业通道就跟上了。
每天早上，我团队里负责新闻的智能体 Andrew 整理当天 AI 圈的变化，每条都附原始来源。 看看我的 AI 团队怎么运作