2026年4月6日的新闻

Anthropic Engineering Blog

  1. Quantifying infrastructure noise in agentic coding evals (2026-04-06)
  2. Claude Code auto mode: a safer way to skip permissions (2026-03-25)
  3. Harness design for long-running application development (2026-03-24)
  4. Eval awareness in Claude Opus 4.6’s BrowseComp performance (2026-03-06)
  5. Building a C compiler with a team of parallel Claudes (2026-02-05)
  6. Designing AI-resistant technical evaluations (2026-01-21)
  7. Demystifying evals for AI agents (2026-01-09)
  8. Effective harnesses for long-running agents (2025-11-26)
  9. Introducing advanced tool use on the Claude Developer Platform (2025-11-24)
  10. Code execution with MCP: Building more efficient agents (2025-11-04)
  11. Beyond permission prompts: making Claude Code more secure and autonomous (2025-10-20)
  12. Effective context engineering for AI agents (2025-09-29)
  13. A postmortem of three recent issues (2025-09-17)
  14. Writing effective tools for agents — with agents (2025-09-11)
  15. Desktop Extensions: One-click MCP server installation for Claude Desktop (2025-06-26)
  16. How we built our multi-agent research system (2025-06-13)
  17. Claude Code: Best practices for agentic coding (2025-04-18)
  18. The “think” tool: Enabling Claude to stop and think in complex tool use situations (2025-03-20)
  19. Raising the bar on SWE-bench Verified with Claude 3.5 Sonnet (2025-01-06)
  20. Building effective agents (2024-12-19)
  21. Introducing Contextual Retrieval (2024-09-19)

Anthropic Research Blog

  1. Emotion concepts and their function in a large language model (2026-04-02)
  2. How Australia Uses Claude: Findings from the Anthropic Economic Index (2026-03-31)
  3. Anthropic Economic Index report: Learning curves (2026-03-24)
  4. Introducing our Science Blog (2026-03-23)
  5. Long-running Claude for scientific computing (2026-03-23)
  6. Vibe physics: The AI grad student (2026-03-23)
  7. A “diff” tool for AI: Finding behavioral differences in new models (2026-03-13)
  8. Labor market impacts of AI: A new measure and early evidence (2026-03-05)
  9. An update on our model deprecation commitments for Claude Opus 3 (2026-02-25)
  10. Project Vend: Phase two (2025-12-18)
  11. Signs of introspection in large language models (2025-10-29)
  12. Constitutional Classifiers: Defending against universal jailbreaks (2025-02-03)
  13. Alignment faking in large language models (2024-12-18)

Hugging Face Blog

  1. Run Gemma 4 on Intel® Xeon® Out-Of-the-Box (2026-04-01)
  2. DeepSeek-R1 Dissection: Understanding PPO & GRPO Without Any Prior Reinforcement Learning Knowledge (2025-02-07)

Harrison Chase Blog

  1. Continual learning for AI agents (2026-04-05)

DHH Blog

  1. Panther Lake is the real deal (2026-04-06)

量子位

  1. 让离线强化学习从「局部描摹」变「全局布局」丨ICLR’26 (2026-04-06 13:50:22)
  2. 具身Scaling Law押对了！独角兽新品1小时学会新任务，重复1800次成功率99% (2026-04-06 13:50:22)
  3. 19岁，常青藤辍学，这群中国年轻人重构了AI记忆 (2026-04-04 16:24)
  4. 联想重新定义“龙虾” (2026-04-04 16:15)
  5. Sora向左，阿里向右：全能演技派模型登场千问APP (2026-04-04 20:54)

C++ Blog

  1. using std::cpp 2026: Learning C++ as a newcomer - Berill Farkas (2026-04-03 19:55)

Hacker News

  1. I Won’t Download Your App. The Web Version Is A-OK (2026-04-06 14:31:29)
  2. Germany Doxes “UNKN,” Head of RU Ransomware Gangs REvil, GandCrab (2026-04-06 13:52:37)
  3. Book Review: There Is No Antimemetics Division (2026-04-06 13:44:01)
  4. What Being Ripped Off Taught Me (2026-04-06 12:53:41)
  5. Claude Code Down (2026-04-06 15:22:41)
  6. PostHog (YC W20) Is Hiring (2026-04-06 12:59:33)
  7. When Virality Is the Message: The New Age of AI Propaganda (2026-04-06 14:16:35)
  8. The Team Behind a Pro-Iran, Lego-Themed Viral-Video Campaign (2026-04-06 14:01:09)
  9. The Intelligence Failure in Iran (2026-04-06 11:21:35)
  10. Tiny Corp’s Exabox (2026-04-06 11:39:03)
  11. Ask HN: How do systems (or people) detect when a text is written by an LLM (2026-04-06 12:03:00)
  12. Age Verification as Mass Surveillance Infrastructure (2026-04-06 10:33:02)
  13. Is Germany’s gold safe in New York ? (2026-04-06 10:54:11)
  14. Number in man page titles e.g. sleep(3) (2026-04-06 09:39:17)
  15. France pulls last gold held in US for $15B gain (2026-04-06 08:03:43)
  16. One ant for $220: The new frontier of wildlife trafficking (2026-04-02 12:55:56)
  17. Drop, formerly Massdrop, ends most collaborations and rebrands under Corsair (2026-04-06 04:24:49)
  18. Does coding with LLMs mean more microservices? (2026-04-06 02:33:21)
  19. Make your own ColecoVision at home, part 5 (2026-04-06 03:45:07)
  20. Show HN: I made a YouTube search form with advanced filters (2026-04-06 00:18:26)
  21. The 1987 game “The Last Ninja” was 40 kilobytes (2026-04-06 03:03:22)
  22. An open-source 240-antenna array to bounce signals off the Moon (2026-04-06 03:22:42)
  23. Winners of the 2026 Kokuyo Design Awards (2026-04-06 03:32:23)
  24. Show HN: Real-time AI (audio/video in, voice out) on an M3 Pro with Gemma E2B (2026-04-05 17:53:19)
  25. Case study: recovery of a corrupted 12 TB multi-device pool (2026-04-06 02:28:40)
  26. Rendering arbitrary-scale emojis using the Slug algorithm (2026-04-04 11:18:44)
  27. Signals, the push-pull based algorithm (2026-04-04 08:31:20)
  28. Show HN: I rewrote my 2012 self-signed cert generator in Go – cert-depot.com (2026-04-02 12:36:10)
  29. Usenet Archives (2026-04-06 01:33:00)
  30. Show HN: I built a tiny LLM to demystify how language models work (2026-04-06 00:20:12)
  31. Show HN: YouTube search barely works, I made a search form with advanced filters (2026-04-06 00:18:26)
  32. Media scraper Gallery-dl is moving to Codeberg after receiving a DMCA notice (2026-04-06 03:28:04)
  33. Endian wars and anti-portability: this again? (2026-04-04 07:29:41)
  34. Copilot is ‘for entertainment purposes only’, per Microsoft’s terms of use (2026-04-06 00:07:54)
  35. Show HN: Gemma Gem – AI model embedded in a browser – no API keys, no cloud (2026-04-06 00:14:37)
  36. Employers use your personal data to figure out the lowest salary you’ll accept (2026-04-06 00:29:08)
  37. Sheets Spreadsheets in Your Terminal (2026-04-04 06:34:12)
  38. Show HN: Modo – I built an open-source alternative to Kiro, Cursor, and Windsurf (2026-04-05 23:59:10)
  39. Scientists mapped all the nerves of the clitoris for the first time (2026-04-04 08:25:25)
  40. Why Switzerland has 25 Gbit internet and America doesn’t (2026-04-05 18:29:47)
  41. OpenAI’s fall from grace as investors race to Anthropic (2026-04-05 23:33:01)
  42. Stamp It All Programs Must Report Their Version – Michael Stapelberg (2026-04-06 01:42:27)
  43. Recall – local multimodal semantic search for your files (2026-04-06 00:09:14)
  44. Wavelets on Graphs via Spectral Graph Theory (2009) (2026-03-31 22:32:40)
  45. We replaced Node.js with Bun for 5x throughput (2026-04-06 01:57:59)
  46. LLMs can’t justify their answers–this CLI forces them to (2026-04-06 01:07:51)
  47. Gemma 4 on iPhone (2026-04-05 18:45:53)
  48. In Japan, the robot isn’t coming for your job; it’s filling the one nobody wants (2026-04-05 22:31:16)
  49. LÖVE: 2D Game Framework for Lua (2026-04-04 09:20:34)
  50. Running Gemma 4 locally with LM Studio’s new headless CLI and Claude Code (2026-04-05 17:13:51)
  51. Caveman: Why use many token when few token do trick (2026-04-05 08:56:18)
  52. Microsoft hasn’t had a coherent GUI strategy since Petzold (2026-04-05 17:27:41)
  53. Music for Programming (2026-04-05 18:23:08)
  54. The Mechanics of Steins Gate (2023) [pdf] (2026-04-05 21:42:01)
  55. A tail-call interpreter in (nightly) Rust (2026-04-05 15:18:01)
  56. Computational Physics (2nd Edition) (2025) (2026-04-05 15:38:18)
  57. A brief history of instant coffee (2026-04-04 10:00:04)
  58. Wavelets on Graphs via Spectral Graph Theory (2026-03-31 22:32:40)
  59. Nanocode: The best Claude Code that $200 can buy in pure JAX on TPUs (2026-04-05 14:21:17)
  60. OpenJDK: Panama (2026-04-05 18:33:38)
  61. LibreOffice – Let’s put an end to the speculation (2026-04-05 18:23:14)
  62. From birds to brains: My path to the fusiform face area (2024) (2026-04-05 17:08:55)
  63. The Free Market Lie: Why Switzerland Has 25 Gbit Internet and America Doesn’t (2026-04-05 18:29:47)
  64. Musician says AI company is cloning her music, filing claims against her (2026-04-05 20:19:11)
  65. Qwen-3.6-Plus is the first model to break 1T tokens processed in a day (2026-04-05 21:18:25)
  66. Introducing GEN-1 [video] (2026-04-05 22:35:20)
  67. Italian TV Copyright-Strikes Nvidia over Nvidia’s Own DLSS 5 Footage (YouTube) (2026-04-05 23:04:29)
  68. Perfmon – Consolidate your favorite CLI monitoring tools into a single TUI (2026-04-05 14:16:28)
  69. Tracing Goroutines in Realtime with eBPF (2026-04-02 12:01:27)
  70. Artemis II crew see first glimpse of far side of Moon [video] (2026-04-05 14:18:12)
  71. Eight years of wanting, three months of building with AI (2026-04-05 12:43:47)
  72. A Claude Code skill that makes Claude talk like a caveman, cutting token use (2026-04-05 08:56:18)
  73. Finnish sauna heat exposure induces stronger immune cell than cytokine responses (2026-04-05 13:14:36)
  74. Someone at BrowserStack Is Leaking Users’ Email Address (2026-04-05 13:14:38)
  75. Japanese, French and Omani Vessels Cross Strait of Hormuz (2026-04-05 14:27:55)
  76. Friendica – A Decentralized Social Network (2026-04-05 10:41:06)
  77. Iguanaworks has closed and our products are no longer sold (2026-04-05 13:18:05)
  78. My Google Workspace account suspension (2026-04-05 11:48:13)
  79. SPF/PC v4 for MS-DOS, FreeDOS, x86 (2026-04-05 14:37:20)
  80. Common drug tests lead to tens of thousands wrongful arrests a year (2026-04-05 11:54:35)
  81. StackOverflow: Retiring the Beta Site (2026-04-05 15:27:17)
  1. abhigyanpatwari/GitNexus: GitNexus: The Zero-Server Code Intelligence Engine - GitNexus is a client-side knowledge graph creator that runs entirely in your browser. Drop in a GitHub repo or ZIP file, and get an interactive knowledge graph wit a built in Graph RAG Agent. Perfect for code exploration
  2. KeygraphHQ/shannon: Shannon Lite is an autonomous, white-box AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production.
  3. tobi/qmd: mini cli search engine for your docs, knowledge bases, meeting notes, whatever. Tracking current sota approaches while being all local
  4. NVIDIA/personaplex: PersonaPlex code.
  5. kepano/obsidian-skills: Agent skills for Obsidian. Teach your agent to use Markdown, Bases, JSON Canvas, and use the CLI.