Skip to content

#开源生态

5 today

Oct 21Wednesday

AI HOT · Industry

ChinaTalk parses MOFCOM Announcement No. 48 and its effect on AI tech export controls

MOFCOM and the Ministry of Science and Technology plan to expand the technology export control catalogue, and ChinaTalk breaks down what Announcement No. 48 means for AI models, chip design, agents and sensitive data technology. The draft adds 40 items: 14 proposed for an export ban and 26 for licensing. Affected technology must be registered before export during the transition period, and comments run until November 15, 2026.

Why it matters: ChinaTalk maps where the proposed rules draw the line on open source and overseas service licensing, which sets different compliance duties for Chinese model firms' weight releases and technology licensing.

Latent Space

DeepSeek open-sources DeepInfer, discloses 155T tokens/day inference on a single cluster with caching

DeepSeek released DeepInfer, the inference engine behind V4.1 Flash, along with deepseek-recipe, the component that handles API requests. The engine supports distributed inference, speculative decoding and structured output. DeepSeek also published production figures showing 155T tokens per day on a single cluster with caching.

Why it matters: The production numbers and cache hit rates let teams estimate inference margins, since cache billing and hardware cost assumptions drive the math.

Hacker News front page

C# now beats Rust on WASM size

NetWasm 0.8.0 编译同一 print-42 工作负载产出 47,798 字节 Wasm,比 Rust 1.95.0 的 53,635 字节小 10.9%,Brotli-11 压缩后小 20.4%。

TechCrunch · AI

Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers

Google 周二宣布开源 TorchTPU,让现有 PyTorch 应用只需修改配置、无需重写代码即可迁移至 TPU 和 Google Cloud。Amazon 周一发布的 TorchNeuron 同样支持应用迁往 Trainium 和 AWS,并向 PyTorch 主项目提交解除部分功能与 Nvidia 芯片绑定的代码。

Hacker News front page

Google open-sources TorchTPU, a native PyTorch back end for TPUs

Google 开源 TorchTPU,为 TPU 提供原生 PyTorch 后端,让 PyTorch 模型和算子可直接在 TPU 设备上运行。项目包含自定义 ATen kernel、XLA 计算编译缓存,支持数据并行与张量并行分布式训练,并集成 torch.compile() 图模式执行。

Hacker News front page

Launch HN: Respan (YC W24) – Open-source Rust AI gateway built around plugins

Respan 发布开源 Rust AI 网关,以小型核心和公开接口支持插件化扩展,目前处于 pre-alpha 阶段。供应商适配、MCP、实时语音、图像与视频,以及智能路由、回退和个人身份信息脱敏均可通过插件实现;跨协议转换并不覆盖所有功能。

Hacker News front page

Show HN: Penmentor – Claude marks up your handwritten math like a teacher

Penmentor 将 Claude、Gemini 或 GPT 接入 xournal++,按键即可在手写数学页面上标注错误、给出提示并生成交互动画。它通过 Claude Code、Codex 或 Antigravity CLI 使用已有订阅,支持整页读取、分级提示和基于 manim 的动画;演示视频使用脚本控制笔迹及固定模型回答。

Oct 20Tuesday

AI HOT · Tips & opinions

Google 发布 TorchTitan TPU 分叉及模型优化指南

Google 发布适配 Google Cloud TPU 的 TorchTitan 分叉,提供分布式训练配方及按步骤执行的模型优化指南。SimpleFSDP、Graph Trainer 和 Pallas 内核支持分片、整步图捕获及计算优化,FLUX.1 案例的训练步耗时从 11.49 秒降至 2.78 秒,达到同硬件 JAX MaxDiffusion 吞吐量的约 99%。

Hacker News front page

Tsrs: A Port of TypeScript to Rust

Cloudflare、VoidZero 与 Owner.com 的 Max Schwenk 合作开发 tsrs,将 TypeScript 编译器移植到 Rust,实验版本已可通过 npm 安装。

Hacker News front page

DeepInfer: High-performance, large-scale inference framework for DeepSeek models

DeepSeek 开源支撑其大规模生产服务的推理框架 DeepInfer,本次提供支持 DeepSeek-V4.1-Flash 的离线推理能力。框架支持优化内核、异步调度、确定性采样和结构化输出,可运行于 NVIDIA SM100 GPUs 与华为昇腾 950 NPUs,采用 Apache License 2.0。

Latent Space

[AINews] not much happened today

Latent Space 汇总 DeepSeek V4.1 Flash 1017、MiniMax M3.1 Flash 和 Mercury 3 的发布信息、定价与评测,并判断它们暂时不会得到广泛使用。

Hacker News front page

Drupal is now light-years ahead of its reputation

Drupal 在 DrupalCon Rotterdam 2026 主题演讲中回顾 CMS、Canvas 与 AI 的进展,并推出 Drupal Advocacy Program,以缩小产品能力与外界认知的差距。

TechCrunch · AI

Amazon is taking its biggest shot at Nvidia yet with new TorchNeuron dev tool

Amazon Web Services 周一发布 TorchNeuron,让使用 PyTorch 编写的 AI 应用更容易迁移到其 Trainium 芯片。Amazon 称只需修改一行配置,无需重写代码,并以 Deepgram 为例称其在生产级准确率下每美元可处理的音频增加了 70%。Amazon 还计划向 PyTorch 上游贡献代码,减少框架与特定芯片的绑定,让开发者拥有更多硬件选择。

Hacker News front page

Android apps running native on macOS

Mac.apk 让 Android 应用以普通 macOS 进程形式在 Apple Silicon Mac 上原生运行,无需模拟器、虚拟机或底层 Linux 内核,应用通过 Mac GPU 绘制并以原生速度执行。

Hacker News front page

I no longer care to work on any Ruby gems so I've archived all my repositories

一名 Ruby gems 维护者宣布归档全部仓库、停止维护,原因是对 Ruby 生态转向 LLM 智能体失去热情。他批评 Rails 负责人使用智能体生成应用,以及 Matz 将 Ruby 定位为面向 AI 时代的编程语言。他认为,人类编写的开源代码正被大型企业用于训练模型,而模型生成的内容又污染了这一公共资源。

Hacker News front page

Launch HN: Laminar (YC S24) – RL-trained model that finds errors in agent traces

Laminar 发布 flow-1,用 RL 训练模型查找智能体轨迹中的错误,并称其在自家基准上检测质量匹配 GPT-6-sol、成本低 23x。模型由小型开源模型微调并经数千次合成调查训练,在 Signals 中检索轨迹片段,按用户定义的 JSON 结构返回发现。

TechCrunch · AI

Nvidia launches new open models to run agents in the enterprise and on PCs

Nvidia 周四发布 Nemotron 3.5 Super VL 和两款轻量 Bolt 开放权重模型,分别面向企业智能体与本地电脑部署。Super VL 拥有 120-billion 参数,支持组织在自有数据上后训练并部署到自有基础设施;Nemotron 3.5 Bolt 74B-A9B 24GB 面向本地多步骤推理、工具调用和网页浏览,Bolt E4B 面向对话智能体。

Oct 19Monday

Hacker News front page

Show HN: KiviDB, a faster Redis-compatible memory database and vector cache

KiviDB 是用 Rust 编写的开源多线程内存数据库,兼容 Redis 协议并内置全文与向量搜索。项目公布的 32 核测试中,KiviDB 1.1.0 的流水线 GET 吞吐为 21.2M ops/s,约为 Redis 8 的 9×、Dragonfly 的 2.3×,测试结果依赖环境与工作负载。

Hacker News front page

I built a programming language that LLMs can't use

开发者构建并开源 KillSwitch,通过对抗性语言设计测试 LLM 和编码智能体能否理解、编写未见过的代码。它是一种基于栈的解释型语言,利用字符计数、规范中的隐藏谎言及触发安全护栏的语句增加难度,示例见 https://killswitch-lang.org 。

Hacker News front page

Show HN: WhatsApp MCP running free on Cloudflare

WhatsApp MCP 可部署在用户自己的 Cloudflare 账户,让 Claude Code、Claude Desktop 等 MCP 客户端读取和搜索 WhatsApp 聊天,但不能发送消息。

Computing Life · Share · Yage

相机、无人机与手机哪些已被破解,AI 在硬件逆向中做了什么

消费设备的解锁涉及官方开放、替代固件、隐藏参数、临时内存补丁和持久扩展五类机制,不能统称为攻破底层安全锁。适马 fp 的 fpSup 属于运行内存补丁,重启即失效;大疆 Mavic 4 Pro 的公开信号增强工具修改遥控器侧参数,未改飞机固件。

Oct 18Sunday

Hacker News front page

Show HN: Sphinx – A hook that quizzes you on your AI's code before shipping

Sphinx 是面向智能体辅助开发的免费开源 Git 钩子,在受保护分支推送前根据代码差异提问,答对后才继续推送。独立审查模型不读取编程聊天记录,回答不会传给编程智能体,也不会打开 .env 文件和密钥;可用 sphinx yolo 跳过检查,每次跳过都会记录。

Oct 17Saturday

The Decoder

DeepSeek-V4.1-Flash-1017 beats larger models on DeepSeek's own agent benchmarks at lower task cost

DeepSeek released DeepSeek-V4.1-Flash-1017, which beats its predecessor and some larger competitors on several of DeepSeek's own coding and agent benchmarks. It is available through the website, app and API, and the model weights are open.

Why it matters: DeepSeek reports cutting cost per task from about $0.50 to under $0.20 on its own coding benchmark while scoring higher accuracy than the previous version.

The Decoder

China drafts export controls on open-weight models and 23 key AI technologies

China's commerce and science ministries released a draft revision of the technology export catalog that would put 23 key AI technologies under export control, covering large models, agents, compute clusters and data processing. A model's core technology is caught if it reaches 10^26 FLOPs of training compute, has at least 2 trillion parameters, or ranks in the global top 10. Models trained on China's "important data" and serving users from that training are also restricted.

Why it matters: China plans licensing thresholds for core large-model technology, with different procedures for overseas local servers and overseas cloud deployments, which directly shapes how models can be deployed abroad.