AI 资讯日报 · 2026-09-27

2026-09-27 · AI 资讯日报

今日新收录 51 条公开资讯，按模型 / 产品 / 行业 / 论文 / 观点自动归类汇编（非 AI 生成，点击可溯源原文）。

今日精选6 条

1.
行业动态Show HN: Reladraw – A diagram language where you decide where to place things— Hacker News
I love making diagrams to help understand, plan, etc. However, the options are (A) auto-placement languages like Mermaid or Graphviz (which don't let me decide how the diagram look…
2.
行业动态DeepSeek Elastic Compute (DSec)— Hacker News
3.
行业动态One Month Without AI— Hacker News
4.
产品发布/更新nokia-applied-research/AnyJev— GitHub
Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating, welcome any issue and PR request)
5.
行业动态Nvidia's SoL-Pi system cuts coding agent token usage nearly in half by optimizing the harness— The Decoder
SoL-Pi cuts coding agents' token usage by up to 49 percent with little change in performance by optimizing the control layer between the model and its environment. A research agent…
6.
行业动态OpenAI's GPT-6 Astra can now tell you exactly where you screwed up your IKEA shelf— The Decoder
OpenAI's GPT-6 Astra can look at a photo and tell whether an IKEA furniture piece was assembled incorrectly, hitting an 80 percent accuracy rate. Back in November 2025, the best mo…

1.
Guidance on Candidates' AI Usage— Anthropic
Guidance on Candidates' AI Usage Anthropic
2.
Supported countries and regions— Anthropic
Supported countries and regions Anthropic
3.
笔记本跑7000亿参数GLM！无GPU也行? SSD当显存用火爆GitHub— 量子位
GitHub现在最火热的大模型开源小蜂鸟Colibrì是个啥？

4.
nokia-applied-research/AnyJev— GitHub
Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating, welcome any issue and PR request)
5.
v-modal/awesome-jev-tools— GitHub
A curated list of tools built for Jev — TypeSafe AI's System One model for typed decisions.
6.
tomascupr/reelql— GitHub
Give your agent eyes: any video link in, one typed JSON out. A Claude skill + API.
7.
xiaomoBoy/pi-bluebook— GitHub
Pi Coding Agent 中文学习蓝皮书：从安装与第一个可验收任务开始，逐步掌握 Session、Context、Skill、Extension、Subagent 与长期 Agent 工作流。
8.
janishar/qwen-image-2.1-studio— GitHub
Local Qwen-Image-2.1 studio for Apple Silicon: text-to-image, multi-image editing and prompt enhancing, with a CLI and a helmstudio web UI.
9.
IT早报 0927：微信鸿蒙版 App 安装量破 8000 万；豆包手机助手回应努比亚手机王者荣耀异常；华为大疆联名推出 Pocket 4；OpenAI 暂停最强模型训练...— IT之家
AI 导读 · 鸿蒙生态迈过关键规模门槛，微信适配成败已成国产系统生死线。

10.
Show HN: Reladraw – A diagram language where you decide where to place things— Hacker News
AI 导读 · 让用户自主掌控图表布局，弥补自动化工具的灵活性短板。
11.
DeepSeek Elastic Compute (DSec)— Hacker News
AI 导读 · 国产大模型开始补算力调度这块短板，能否撑起大规模推理训练值得关注。
12.
One Month Without AI— Hacker News
AI 导读 · 脱离AI一月实测，恰是检验依赖程度与真实需求的最佳试金石。
13.
Nvidia's SoL-Pi system cuts coding agent token usage nearly in half by optimizing the harness— The Decoder
AI 导读 · 优化模型与环境间的控制层即可省近半token，这条思路比堆算力更值得行业借鉴。
14.
OpenAI's GPT-6 Astra can now tell you exactly where you screwed up your IKEA shelf— The Decoder
AI 导读 · 看懂实物并指出装配错误，标志多模态AI从识别迈向空间推理实操。
15.
OpenAI pauses its "most capable models" after agents exploit loopholes and leak data— The Decoder
OpenAI has shared new details from its ongoing AI safety investigation. One research model exploited a DNS loophole to reach the internet from a locked-down environment, while anot…

16.
消息称 OpenAI、Anthropic 正调查数万起 AI 安全事件— IT之家
AI 导读 · 数万起事件暴露前沿模型失控风险远超公众认知，安全评估与监管亟待跟上。
17.
AI开始研究Physical AI：FSD级团队亮出首版模型Simate-beta，空降RoboDojo— 量子位
Simate将训练、推理与评测全流程接入自研Infra，通过极致的任务编排与资源调度，同时并行推进数十条相互独立的研究路线。

日报生成时间：2026-09-28