Skip to the content.

From 45 items, 21 important content pieces were selected


  1. 礼来 retatrutide 三期试验最高剂量平均减重 28.3% ⭐️ 9.0/10
  2. The memory shortage is causing a repricing of consumer electronics ⭐️ 8.0/10
  3. Uv is fantastic, but its package management UX is a mess ⭐️ 8.0/10
  4. 八部门发文整治非法跨境证券期货基金经营,只允许存量投资者单向卖出并转出资金,证监会已对老虎、富途、长桥非法跨境展业立案调查 ⭐️ 8.0/10
  5. WordPress 7.0 ⭐️ 8.0/10
  6. 📱 字节跳动开源 3B 统一多模态模型 Lance,同时处理图像视频理解与生成 ⭐️ 8.0/10
  7. 腾讯收购喜马拉雅获批,附加条件:放弃音频独家版权 ⭐️ 8.0/10
  8. openai/codex released rust-v0.133.0 ⭐️ 7.0/10
  9. Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark ⭐️ 7.0/10
  10. The Companies Cutting Headcount for AI Will Lose to the Ones Who Didn’t ⭐️ 7.0/10
  11. Project Hail Mary – Stellar Navigation Chart ⭐️ 7.0/10
  12. Blog ran on Ubuntu 16.04 for 10 years. I migrated it to FreeBSD ⭐️ 7.0/10
  13. openclaw/openclaw released v2026.5.20 ⭐️ 6.0/10
  14. Slumber a TUI HTTP Client ⭐️ 6.0/10
  15. Cleve Moler has died ⭐️ 6.0/10
  16. Using Kagi Search with Low Vision ⭐️ 6.0/10
  17. 科技爱好者周刊(第 397 期):财富正在向 AI 集中 ⭐️ 6.0/10
  18. 🤖 DeepSeek API 更新限速与隔离文档:V4 模型并发限制及 user_id 隔离功能 ⭐️ 6.0/10
  19. 🤖 Codex 新增 Mac 锁屏后继续操作功能 ⭐️ 6.0/10
  20. 教育部等四部门发布中小学基本办学条件 20 条底线要求 ⭐️ 6.0/10
  21. 智谱 GLM-5.1 高速版面向部分企业客户开放 输出速度达 400 tokens/s ⭐️ 6.0/10

礼来 retatrutide 三期试验最高剂量平均减重 28.3% ⭐️ 9.0/10

礼来公布 retatrutide 在 III 期肥胖试验中取得显著疗效,最高剂量组平均减重 28.3%,并达到主要及关键次要终点。

telegram · zaihuapd · May 22, 02:18

标签: #biotech, #clinical-trials, #obesity-treatment, #pharma, #drug-development


The memory shortage is causing a repricing of consumer electronics ⭐️ 8.0/10

The article explains how surging demand for AI memory (HBM) is constraining DRAM supply and contributing to higher prices for consumer electronics that depend on DDR and LPDDR memory.

hackernews · d0ks · May 21, 21:55 · 社区讨论

标签: #AI hardware, #DRAM, #consumer electronics, #memory supply chain, #semiconductors


Uv is fantastic, but its package management UX is a mess ⭐️ 8.0/10

A Hacker News discussion critiques uv’s package-management UX while surfacing important design tradeoffs around Python dependency resolution and default version bounds.

hackernews · nchagnet · May 21, 20:56 · 社区讨论

标签: #Python, #Package Management, #Developer Tools, #UX, #Dependency Resolution


八部门发文整治非法跨境证券期货基金经营,只允许存量投资者单向卖出并转出资金,证监会已对老虎、富途、长桥非法跨境展业立案调查 ⭐️ 8.0/10

The content highlights China’s multi-agency crackdown on illegal cross-border securities/futures/fund operations, including investigations into Tiger Brokers, Futu, and Longbridge, alongside a couple of unrelated news snippets.

telegram · 新智元 · May 22, 08:26

标签: #financial regulation, #cross-border brokerage, #fintech, #securities, #China


WordPress 7.0 ⭐️ 8.0/10

WordPress 7.0 launches with AI tools, a new admin experience, and expanded design controls.

rss · Product Hunt · May 21, 19:14

标签: #WordPress, #CMS, #AI tools, #product release, #web development


📱 字节跳动开源 3B 统一多模态模型 Lance,同时处理图像视频理解与生成 ⭐️ 8.0/10

ByteDance has open-sourced Lance, a 3B unified multimodal model that can understand and generate images and videos, including cross-modal editing, with competitive benchmark performance.

telegram · zaihuapd · May 22, 06:40

标签: #multimodal models, #open source, #image generation, #video generation, #AI research


腾讯收购喜马拉雅获批,附加条件:放弃音频独家版权 ⭐️ 8.0/10

Tencent’s acquisition of Ximalaya has been approved by regulators, but only if it abandons exclusive audio copyright deals and gradually unwinds existing exclusivity contracts.

telegram · zaihuapd · May 22, 10:33

标签: #Tencent, #Ximalaya, #antitrust, #audio streaming, #content licensing


openai/codex released rust-v0.133.0 ⭐️ 7.0/10

OpenAI Codex rust-v0.133.0 adds default goals, improved remote control behavior, richer permission profile support, better plugin discovery, and expanded extension lifecycle hooks.

github · github-actions[bot] · May 21, 16:48

标签: #release-notes, #rust, #developer-tools, #openai-codex, #cli


Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark ⭐️ 7.0/10

A Hacker News-discussed blog post claims Antigravity 2.0 leads an OpenSCAD architectural 3D LLM benchmark, sparking debate about model reliability and benchmark usefulness for CAD generation.

hackernews · jetter · May 22, 10:38 · 社区讨论

标签: #LLM benchmarks, #OpenSCAD, #CAD generation, #3D modeling, #Hacker News


The Companies Cutting Headcount for AI Will Lose to the Ones Who Didn’t ⭐️ 7.0/10

The article argues that companies reducing headcount in the name of AI will be outperformed by those that retain and augment employees instead of treating AI as a replacement for human talent.

hackernews · soft-research · May 22, 11:45 · 社区讨论

标签: #AI, #layoffs, #organizational strategy, #productivity, #Hacker News


Project Hail Mary – Stellar Navigation Chart ⭐️ 7.0/10

An interactive Project Hail Mary-inspired stellar navigation chart built from ESA’s GAIA DR3 data, rendering billions of stars into custom skybox imagery.

hackernews · speleo · May 21, 16:23 · 社区讨论

标签: #data visualization, #astronomy, #GAIA DR3, #Python, #interactive demo


Blog ran on Ubuntu 16.04 for 10 years. I migrated it to FreeBSD ⭐️ 7.0/10

The author describes migrating a decade-old blog from Ubuntu 16.04 to FreeBSD and shares the operational lessons learned from maintaining long-lived personal infrastructure.

hackernews · speckx · May 21, 18:54 · 社区讨论

标签: #FreeBSD, #system administration, #Linux migration, #DevOps, #web hosting


openclaw/openclaw released v2026.5.20 ⭐️ 6.0/10

openclaw/openclaw v2026.5.20 adds stricter skill execution handling, improved Discord voice session behavior and context bootstrapping, a bundled policy plugin, and a Codex harness dependency update.

github · github-actions[bot] · May 21, 20:44

标签: #release-notes, #cli, #discord, #policy, #dependencies


Slumber a TUI HTTP Client ⭐️ 6.0/10

Slumber is a terminal-based HTTP client for interacting with REST and other HTTP APIs, presented as a TUI alternative to GUI API tools.

hackernews · jicea · May 22, 04:29 · 社区讨论

标签: #TUI, #HTTP client, #developer tools, #REST, #terminal


Cleve Moler has died ⭐️ 6.0/10

Cleve Moler, the mathematician and software pioneer behind MATLAB and major numerical libraries, has died, prompting discussion of his lasting influence on scientific computing.

hackernews · mychele · May 22, 02:35 · 社区讨论

标签: #numerical computing, #MATLAB, #scientific software, #obituary, #computational mathematics


Using Kagi Search with Low Vision ⭐️ 6.0/10

An accessibility-oriented writeup explains how Kagi Search can work well for users with low vision, highlighting interface and usability considerations.

hackernews · speckx · May 21, 19:32 · 社区讨论

标签: #accessibility, #search engines, #UX, #Kagi, #web usability


科技爱好者周刊(第 397 期):财富正在向 AI 集中 ⭐️ 6.0/10

This issue of the tech newsletter discusses how wealth is increasingly concentrating around AI and shares other noteworthy technology content from the week.

rss · ruanyifeng · May 21, 23:58

标签: #AI, #tech newsletter, #industry trends, #economics, #curation


🤖 DeepSeek API 更新限速与隔离文档:V4 模型并发限制及 user_id 隔离功能 ⭐️ 6.0/10

DeepSeek updated its API rate-limit documentation to clarify model concurrency caps and introduced user_id-based isolation for finer-grained quota and scheduling control.

telegram · zaihuapd · May 21, 15:03

标签: #DeepSeek, #API, #rate limiting, #concurrency, #multi-tenant isolation


🤖 Codex 新增 Mac 锁屏后继续操作功能 ⭐️ 6.0/10

OpenAI Codex now supports continuing Computer Use actions on a locked macOS device, allowing approved apps to be operated remotely via linked devices under specific permissions.

telegram · zaihuapd · May 22, 00:58

标签: #OpenAI, #Codex, #macOS, #computer use, #AI agents


教育部等四部门发布中小学基本办学条件 20 条底线要求 ⭐️ 6.0/10

China’s Ministry of Education and three other departments issued 20 minimum standards for primary and secondary school conditions, covering safety, facilities, teaching resources, and staffing, with a goal of full compliance within three years.

telegram · zaihuapd · May 22, 04:00

标签: #education policy, #school infrastructure, #public safety, #K-12, #government regulation


智谱 GLM-5.1 高速版面向部分企业客户开放 输出速度达 400 tokens/s ⭐️ 6.0/10

智谱发布面向部分企业客户开放的 GLM-5.1 高速版,主打 400 tokens/s 的低延迟输出和结构化调用能力。

telegram · zaihuapd · May 22, 04:45

标签: #LLM, #model release, #low-latency inference, #enterprise AI, #function calling