AI 日报 · 2026年9月2日

14 Claude Code Commands That Actually Save Time (Fact Checked)

The Claude Code habits that actually save time, and the outdated advice you should stop followingContinue reading on Tow…

来源:Towards AI

My local model setup on an M4 Pro Mac Mini

Article URL: https://lws.io/blog/my-local-model-setup/ Comments URL: https://news.ycombinator.com/item?id=49529132 Point…

来源:Hacker首页帖子

[分享发现] 原来网页版 codex 才是满血的 5.6 sol

应该是 IP 不干净，用的是 5.6 sol ，实际上是 luna 。大半夜的，突然悟了... 网页版 codex 是满血版，让他做好计划到本地用 luna 跑 你不是能掺水吗，那我直接 luna max 好吧... 前情提要： 应该是昨天…

来源:V2EX

[Cowork/claude.ai] Cannot add anthropics/claude-plugins-community as a marketplace; community plugins (e.g. eli5) missing from the directory

Summary Community plugins such as eli5 can be installed via Claude Code CLI, but there is no working path to install the…

来源:Claude Code（GitHub Issues · RSSHub）

[BUG] Artifact fails to load intermittently — "Connection reset" error

Preflight Checklist [x] I have searched existing issues and this hasn't been reported yet [x] This is a single bug repor…

来源:Claude Code（GitHub Issues · RSSHub）

Claude Fable 5.1 and Claude Mythos 5.1

What's new in Claude Fable 5.1 – https://platform.claude.com/docs/en/models/fable-5-1/whats-n...System Card: https://www…

来源:Hacker News 热门

Claude Fable 5.1 在 GitHub Copilot 中正式上线

Anthropic 旗下 Mythos 系列的 Claude Fable 5.1 模型现已在 GitHub Copilot 中正式可用，擅长长时间跨度自主编码与知识工作类任务。

来源:GitHub Copilot

Open AI’s Astra model is on the way — and very good at breaking into computer systems

OpenAI previewed the precautions it is taking as it prepares to release Astra, its newest, cyber-critical LLM.

来源:TechCrunch

Improving our alignment and security efforts

Article URL: https://www.anthropic.com/news/improving-alignment-security-efforts Comments URL: https://news.ycombinator.…

来源:Hacker首页帖子

AfterQuery reportedly becomes Y Combinator’s fastest-ever unicorn, now valued at $3.2B

AI model-training startup AfterQuery has reportedly raised a round that valued it at $3.2 billion, just five months afte…

来源:TechCrunch

Anthropic发布Claude Fable 5.1：缓存成本降至1/4，长程智能体能力升级

Anthropic推出Claude Fable 5.1，缓存读取成本降至原价四分之一，专为长程智能体编程、多步研究和复杂文档处理优化。新增effort参数动态调整、C2PA内容凭证等功能，含三项破坏性变更。

来源:Hacker News 中文摘要 - 每日 AI 论文精选 | Zeli

[推广] 整理了免费的大模型 API 的靠谱渠道,包括 DeepSeek V4， GLM 5.x 等 30+ 模型

免费的大模型 API 的靠谱渠道,包括 DeepSeek V4 ，GLM 5.x 等 30+ 模型： https://github.com/peter123023/awesome-free-llm-api 收录标准 只收录能拿 API Ke…

来源:V2EX

siyuan-note/siyuan

An open-source, privacy-first, self-hosted knowledge workspace where humans and AI agents work together 开源、隐私优先、自托管的知识工作…

来源:GitHub TypeScript

Waishnav/devspace

Minimal Coding Agent Harness over MCP for ChatGPT, Claude, Hermes, Grok Bot, OpenClaw DevSpace Bring a Codex-style codin…

来源:GitHub TypeScript

微软发布 Phi-4-mini-instruct 指令微调模型

微软在 HuggingFace 上架 Phi-4-mini-instruct 小型指令微调模型，主打轻量高效，可在边缘和低资源场景部署。

来源:HuggingFace Trending Models

every-app/open-seo

Open source alternative to Semrush and Ahrefshttps://openseo.soOpenSEO Open source alternative to Semrush and Ahrefs Ope…

来源:GitHub TypeScript

PurpleDoubleD/locally-uncensored

Plug-and-play local AI studio: uncensored chat, image & video generation, coding agent. Runs abliterated LLMs + Comf…

来源:GitHub TypeScript

langgenius/dify

Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cl…

来源:GitHub TypeScript

Dokploy/dokploy

Open Source Alternative to Vercel, Netlify and Heroku.https://dokploy.com/ Join us on Discord for help, feedback, and di…

来源:GitHub TypeScript

stepfun-ai/stepvideo-t2v

来源:HuggingFace Trending Models

大模型的逐字生成模式，正在拖慢你的Agent，有人跑出了5倍提速

大模型的逐字生成模式，正在拖慢你的Agent，有人跑出了5倍提速

来源:机器之心

AI能挖漏洞吗？北航等机构把真实漏洞炼进了模型参数

AI能挖漏洞吗？北航等机构把真实漏洞炼进了模型参数

来源:机器之心

Why Frontier LLMs Fail to Judge AI Research: Inside Anthropic’s TASTE Benchmark

Anthropic’s TASTE benchmark shows frontier LLMs struggle to judge research proposals, peaking at 60% compared to…

来源:Towards AI

BenchMIRT: What are LLM benchmarks actually measuring?

来源:Hugging Face Blog