{"id":393,"date":"2026-05-11T07:24:56","date_gmt":"2026-05-10T23:24:56","guid":{"rendered":"http:\/\/www.faiyi.com\/?p=393"},"modified":"2026-05-11T07:24:56","modified_gmt":"2026-05-10T23:24:56","slug":"ai%e5%8a%a8%e6%80%81%e6%af%8f%e6%97%a5%e7%ae%80%e6%8a%a5-2026-05-11","status":"publish","type":"post","link":"http:\/\/www.faiyi.com\/?p=393","title":{"rendered":"AI\u52a8\u6001\u6bcf\u65e5\u7b80\u62a5 2026-05-11"},"content":{"rendered":"

\u65e5\u671f\uff1a2026-05-11<\/p>\n

\u672c\u671f\u805a\u7126\uff1a\u91cd\u70b9\u5173\u6ce8\u6a21\u578b\u53d1\u5e03\u4e0e release notes\u3001\u5b98\u65b9 engineering blog\u3001AI coding \/ agent \/ SRE\u3001\u8bc4\u6d4b\u699c\u5355\u53d8\u5316\u3001\u5f00\u53d1\u8005\u5b9e\u8df5\u535a\u5ba2\u3001\u6846\u67b6\u751f\u6001\u3001\u5f00\u6e90\u6a21\u578b\u4e0e\u771f\u5b9e\u7528\u6237\u89c6\u89d2\uff1b\u5f53 HN\u3001Reddit\u3001Hugging Face \u7b49\u793e\u533a\u6e90\u53ef\u8bbf\u95ee\u65f6\u4f18\u5148\u7eb3\u5165\u3002<\/p>\n

\n
    \n
  1. \n

    Artificial Analysis \u6700\u65b0\u6a21\u578b\u6392\u540d\u89c2\u5bdf<\/strong>\uff08Artificial Analysis\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Artificial Analysis \u5e73\u53f0\u6301\u7eed\u8ffd\u8e2a\u5168\u7403\u4e3b\u6d41\u5927\u6a21\u578b\u7684\u667a\u80fd\u6307\u6570\u3001\u8f93\u51fa\u901f\u5ea6\u3001\u5ef6\u8fdf\u3001\u4ef7\u683c\u53ca\u4e0a\u4e0b\u6587\u7a97\u53e3\u7b49\u591a\u7ef4\u6307\u6807\u3002\u6700\u65b0 Intelligence Index v4.0 \u663e\u793a\uff0cGPT-5.5 (xhigh) \u4ee5 60 \u5206\u4f4d\u5c45\u699c\u9996\uff0cClaude Opus 4.7 (Max Effort) \u4e0e Gemini 3.1 Pro Preview \u5e76\u5217\u7b2c\u4e09\uff0857 \u5206\uff09\u3002\u5f00\u6e90\u6a21\u578b\u4e2d\uff0cKimi K2.6 \u4ee5 54 \u5206\u9886\u5148\u3002\u5e73\u53f0\u8fd8\u65b0\u589e\u4e86\u7f13\u5b58\u547d\u4e2d\u5b9a\u4ef7\uff08Cache Hit Price\uff09\u5bf9\u6bd4\uff0c\u5e2e\u52a9\u5f00\u53d1\u8005\u7efc\u5408\u8bc4\u4f30\u6210\u672c\u4e0e\u6027\u80fd\u3002<\/p>\n

    English Summary:<\/strong> Artificial Analysis tracks comprehensive metrics for major AI models including intelligence, speed, latency, pricing, and context windows. The latest Intelligence Index v4.0 ranks GPT-5.5 (xhigh) first with a score of 60, while Claude Opus 4.7 (Max Effort) and Gemini 3.1 Pro Preview tie for third at 57. Among open-weight models, Kimi K2.6 leads with 54. The platform now includes cache hit pricing comparisons to help developers evaluate cost-performance tradeoffs.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  2. \n

    Introducing Claude Opus 4.7<\/strong>\uff08Anthropic News\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Anthropic \u6b63\u5f0f\u53d1\u5e03 Claude Opus 4.7\uff0c\u5728\u9ad8\u7ea7\u8f6f\u4ef6\u5de5\u7a0b\u4efb\u52a1\u4e0a\u8f83 4.6 \u6709\u663e\u8457\u63d0\u5347\uff0c\u5c24\u5176\u5728\u590d\u6742\u957f\u7a0b\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u66f4\u5f3a\u7684\u4e25\u8c28\u6027\u548c\u4e00\u81f4\u6027\u3002\u6a21\u578b\u652f\u6301\u66f4\u9ad8\u5206\u8fa8\u7387\u56fe\u50cf\u8f93\u5165\uff08\u957f\u8fb9\u53ef\u8fbe 2576 \u50cf\u7d20\uff09\uff0c\u5e76\u5728\u4e13\u4e1a\u4efb\u52a1\u4e2d\u5c55\u73b0\u51fa\u66f4\u4f73\u7684\u5ba1\u7f8e\u4e0e\u521b\u9020\u529b\u3002\u5b9a\u4ef7\u7ef4\u6301\u4e0d\u53d8\uff1a\u8f93\u5165 5 \u7f8e\u5143\/\u767e\u4e07 token\uff0c\u8f93\u51fa 25 \u7f8e\u5143\/\u767e\u4e07 token\u3002\u65b0\u589e xhigh \u52aa\u529b\u7ea7\u522b\uff0cClaude Code \u9ed8\u8ba4\u63d0\u5347\u81f3 xhigh\u3002\u540c\u65f6\u63a8\u51fa Cyber Verification Program\uff0c\u4f9b\u5b89\u5168\u4e13\u4e1a\u4eba\u5458\u7533\u8bf7\u5408\u6cd5\u7f51\u7edc\u5b89\u5168\u7528\u9014\u3002<\/p>\n

    English Summary:<\/strong> Anthropic released Claude Opus 4.7, delivering notable improvements in advanced software engineering over 4.6, with stronger rigor and consistency on complex, long-running tasks. The model supports higher-resolution image inputs (up to 2,576 pixels on the long edge) and demonstrates better taste and creativity in professional tasks. Pricing remains unchanged at $5\/M input and $25\/M output tokens. A new xhigh effort level is introduced, with Claude Code defaulting to xhigh.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  3. \n

    Featured An update on recent Claude Code quality reports<\/strong>\uff08Anthropic Engineering\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Anthropic \u5de5\u7a0b\u56e2\u961f\u53d1\u5e03 Claude Code \u8fd1\u671f\u8d28\u91cf\u95ee\u9898\u7684\u590d\u76d8\u62a5\u544a\uff0c\u786e\u8ba4\u4e09\u5904\u72ec\u7acb\u53d8\u66f4\u5bfc\u81f4\u7528\u6237\u4f53\u9a8c\u4e0b\u964d\uff1a3 \u6708 4 \u65e5\u5c06\u9ed8\u8ba4\u63a8\u7406\u52aa\u529b\u7ea7\u522b\u4ece high \u964d\u81f3 medium\uff1b3 \u6708 26 \u65e5\u7684\u7f13\u5b58\u4f18\u5316 bug \u5bfc\u81f4\u4f1a\u8bdd\u8d85\u8fc7\u4e00\u5c0f\u65f6\u95f2\u7f6e\u540e\u6301\u7eed\u4e22\u5931\u5386\u53f2\u601d\u8003\u5185\u5bb9\uff1b4 \u6708 16 \u65e5\u65b0\u589e\u7684\u7cfb\u7edf\u63d0\u793a\u8bcd\u9650\u5236\u8f93\u51fa\u957f\u5ea6\uff0c\u610f\u5916\u5f71\u54cd\u7f16\u7801\u8d28\u91cf\u3002\u6240\u6709\u95ee\u9898\u5df2\u4e8e 4 \u6708 20 \u65e5\u4fee\u590d\uff0c\u56e2\u961f\u627f\u8bfa\u52a0\u5f3a\u5185\u90e8\u6d4b\u8bd5\u6d41\u7a0b\uff0c\u5305\u62ec\u6269\u5927\u5458\u5de5\u4f7f\u7528\u516c\u5f00\u7248\u672c\u7684\u6bd4\u4f8b\u3001\u6539\u8fdb Code Review \u5de5\u5177\uff0c\u5e76\u5bf9\u7cfb\u7edf\u63d0\u793a\u8bcd\u53d8\u66f4\u5b9e\u65bd\u66f4\u4e25\u683c\u7684\u8bc4\u4f30\u4e0e\u6e10\u8fdb\u5f0f\u53d1\u5e03\u3002<\/p>\n

    English Summary:<\/strong> Anthropic's engineering team published a postmortem on recent Claude Code quality issues, identifying three separate changes that caused degraded user experience: lowering default reasoning effort from high to medium on March 4; a caching optimization bug on March 26 that continuously dropped prior reasoning after sessions were idle for over an hour; and a system prompt change on April 16 to reduce verbosity that inadvertently hurt coding quality. All issues were resolved by April 20. The team committed to improving internal testing, expanding staff use of public builds, enhancing Code Review tools, and implementing stricter evaluation and gradual rollouts for system prompt changes.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  4. \n

    Scaling Managed Agents: Decoupling the brain from the hands<\/strong>\uff08Anthropic Engineering\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Anthropic \u5de5\u7a0b\u535a\u5ba2\u6df1\u5165\u4ecb\u7ecd Managed Agents \u67b6\u6784\u8bbe\u8ba1\u54f2\u5b66\uff0c\u6838\u5fc3\u601d\u8def\u662f"\u89e3\u8026\u5927\u8111\u4e0e\u53cc\u624b"\u2014\u2014\u5c06\u4ee3\u7406\u7684\u4f1a\u8bdd\uff08session\uff09\u3001 harness\uff08\u8c03\u7528\u5faa\u73af\uff09\u548c\u6c99\u7bb1\uff08sandbox\uff09\u62bd\u8c61\u4e3a\u72ec\u7acb\u63a5\u53e3\uff0c\u4f7f\u5404\u7ec4\u4ef6\u53ef\u72ec\u7acb\u6f14\u8fdb\u548c\u66ff\u6362\u3002\u8be5\u8bbe\u8ba1\u89e3\u51b3\u4e86\u65e9\u671f\u5355\u4f53\u5bb9\u5668\u67b6\u6784\u7684"\u5ba0\u7269\u670d\u52a1\u5668"\u95ee\u9898\uff0c\u5c06 harness \u79fb\u51fa\u5bb9\u5668\u540e\uff0cp50 \u9996 token \u5ef6\u8fdf\u964d\u4f4e\u7ea6 60%\uff0cp95 \u964d\u4f4e\u8d85 90%\u3002\u901a\u8fc7\u5c06\u51ed\u8bc1\u4e0e\u6c99\u7bb1\u9694\u79bb\u3001\u4f1a\u8bdd\u65e5\u5fd7\u6301\u4e45\u5316\u5b58\u50a8\u5728\u5916\u90e8\uff0c\u7cfb\u7edf\u5b9e\u73b0\u4e86\u66f4\u597d\u7684\u5b89\u5168\u6027\u3001\u53ef\u6062\u590d\u6027\u548c\u6269\u5c55\u6027\uff0c\u652f\u6301\u591a\u5927\u8111\u8fde\u63a5\u591a\u53cc\u624b\u7684\u7075\u6d3b\u90e8\u7f72\u6a21\u5f0f\u3002<\/p>\n

    English Summary:<\/strong> Anthropic's engineering blog details the Managed Agents architecture, centered on decoupling the "brain" (harness) from the "hands" (sandboxes) and session logs. By abstracting sessions, harnesses, and sandboxes into independent interfaces, each component can evolve and be replaced separately. This solved the "pet server" problem of the earlier monolithic container design. Moving the harness out of containers reduced p50 time-to-first-token latency by ~60% and p95 by over 90%.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  5. \n

    Get ready for the whisper-filled office of the future<\/strong>\uff08TechCrunch AI\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>\u968f\u7740\u8bed\u97f3\u8f93\u5165\u4e0e AI \u7f16\u7a0b\u5de5\u5177\u7ed3\u5408\u65e5\u76ca\u666e\u53ca\uff0c\u529e\u516c\u5ba4\u58f0\u5b66\u73af\u5883\u6b63\u5728\u53d1\u751f\u53d8\u5316\u3002\u300a\u534e\u5c14\u8857\u65e5\u62a5\u300b\u62a5\u9053\u6307\u51fa\uff0cWispr \u7b49\u8bed\u97f3\u542c\u5199\u5e94\u7528\u4e0e vibe coding \u5de5\u5177\u7ed3\u5408\u540e\uff0c\u8d8a\u6765\u8d8a\u591a\u5f00\u53d1\u8005\u5728\u529e\u516c\u5ba4\u4f4e\u58f0\u5bf9\u7535\u8111\u8bf4\u8bdd\u3002Gusto \u8054\u5408\u521b\u59cb\u4eba Edward Kim \u9884\u6d4b\u672a\u6765\u529e\u516c\u5ba4\u5c06\u66f4\u50cf\u9500\u552e\u697c\u5c42\uff0c\u4ed6\u672c\u4eba\u5df2\u51e0\u4e4e\u5b8c\u5168\u8f6c\u5411\u8bed\u97f3\u8f93\u5165\u3002\u5c3d\u7ba1\u76ee\u524d\u540c\u4e8b\u95f4\u4f4e\u58f0\u5de5\u4f5c\u7565\u663e\u5c34\u5c2c\uff0c\u4f46 Wispr \u521b\u59cb\u4eba Tanay Kothari \u8ba4\u4e3a\u8fd9\u79cd\u73b0\u8c61\u7ec8\u5c06\u50cf\u4eba\u4eec\u957f\u65f6\u95f4\u76ef\u7740\u624b\u673a\u4e00\u6837\u53d8\u5f97\u5e73\u5e38\u3002<\/p>\n

    English Summary:<\/strong> As voice input combined with AI coding tools becomes more prevalent, office acoustics are shifting. A Wall Street Journal feature highlights the rising popularity of dictation apps like Wispr, especially when connected to vibe coding tools. Gusto co-founder Edward Kim predicts offices will sound more like sales floors in the future, noting he now types only when absolutely necessary.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  6. \n

    MySQL 9.7: First Major LTS Since 8.4 Brings Enterprise Features to Community Edition<\/strong>\uff08InfoQ AI\/ML\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Oracle \u6b63\u5f0f\u53d1\u5e03 MySQL 9.7.0\uff0c\u8fd9\u662f\u81ea MySQL 8.4 \u4ee5\u6765\u7684\u9996\u4e2a\u91cd\u5927 LTS \u7248\u672c\u3002\u65b0\u7248\u672c\u5c06\u591a\u9879\u4f01\u4e1a\u7ea7\u529f\u80fd\u5f00\u653e\u7ed9\u793e\u533a\u7248\uff0c\u5305\u62ec\u539f\u751f\u5411\u91cf\u5b58\u50a8\u3001AI \u9a71\u52a8\u7684\u67e5\u8be2\u4f18\u5316\u5668\u589e\u5f3a\u3001\u6539\u8fdb\u7684 JSON \u5904\u7406\u4ee5\u53ca\u66f4\u5f3a\u7684\u5b89\u5168\u5ba1\u8ba1\u80fd\u529b\u30029.7 LTS \u7cfb\u5217\u627f\u8bfa\u63d0\u4f9b\u957f\u8fbe 8 \u5e74\u7684\u652f\u6301\u5468\u671f\uff0c\u65e8\u5728\u5e2e\u52a9\u4f01\u4e1a\u5728\u4fdd\u6301\u5f00\u6e90\u7075\u6d3b\u6027\u7684\u540c\u65f6\u83b7\u5f97\u4f01\u4e1a\u7ea7\u53ef\u9760\u6027\u3002\u6b64\u6b21\u53d1\u5e03\u6807\u5fd7\u7740 MySQL \u5728\u4e91\u539f\u751f\u548c AI \u5de5\u4f5c\u8d1f\u8f7d\u652f\u6301\u65b9\u9762\u8fc8\u51fa\u4e86\u91cd\u8981\u4e00\u6b65\u3002<\/p>\n

    English Summary:<\/strong> Oracle announced the general availability of MySQL 9.7.0, the first major LTS release since MySQL 8.4. The new version brings enterprise-grade features to the Community Edition, including native vector storage, AI-enhanced query optimizer, improved JSON handling, and enhanced security auditing. The 9.7 LTS series offers an 8-year support lifecycle, enabling enterprises to maintain open-source flexibility while gaining enterprise reliability. This release marks a significant step forward in cloud-native and AI workload support for MySQL.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  7. \n

    [AINews] Anthropic growing 10x\/year while everyone else is laying off >10% of their workforce<\/strong>\uff08Latent Space\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>\u636e\u4e8c\u7ea7\u5e02\u573a\u53ca\u5a92\u4f53\u62a5\u9053\uff0cAnthropic \u5728\u7ecf\u5386\u7b2c\u4e00\u5b63\u5ea6\u5e74\u5316\u589e\u957f 80 \u500d\u7684"\u5947\u8ff9\u5b63\u5ea6"\u540e\uff0c\u4f30\u503c\u5df2\u8fbe 1-1.2 \u4e07\u4ebf\u7f8e\u5143\uff0c\u6b63\u5f0f\u8d85\u8d8a OpenAI \u6210\u4e3a\u5168\u7403\u7b2c 11-15 \u5927\u6700\u6709\u4ef7\u503c\u516c\u53f8\u3002\u4e0e\u6b64\u540c\u65f6\uff0cBlock\u3001Coinbase\u3001Cloudflare \u7b49\u591a\u5bb6\u79d1\u6280\u516c\u53f8\u88c1\u5458 10-40%\uff0c\u5747\u58f0\u79f0\u662f\u4e3a AI \u8f6c\u578b\u505a\u51c6\u5907\u3002\u8fd9\u79cd\u9c9c\u660e\u5bf9\u6bd4\u5f15\u53d1\u4e86\u5bf9 AI \u7ecf\u6d4e\u5206\u5316\u7684\u8ba8\u8bba\uff1a\u771f\u6b63\u53d7\u76ca\u4e8e AI \u7684\u516c\u53f8\u5728\u6269\u5f20\uff0c\u800c\u4f20\u7edf\u8f6f\u4ef6\u516c\u53f8\u5219\u5728\u6536\u7f29\uff0c\u7ecf\u6d4e\u6b63\u5448\u73b0\u5411 AI \u786c\u4ef6\u548c\u80fd\u6e90\u9886\u57df\u9ad8\u5ea6\u96c6\u4e2d\u7684\u8d8b\u52bf\u3002<\/p>\n

    English Summary:<\/strong> Reports from secondary markets and traditional media indicate that Anthropic, following its "miracle Q1" with 80x annualized growth, is now valued at $1-1.2 trillion, officially overtaking OpenAI as the 11th-15th most valuable company globally. In stark contrast, tech companies including Block (40%), Coinbase (14%), and Cloudflare (20%) have laid off significant portions of their workforce, all citing AI readiness. This dichotomy highlights a growing economic divergence: companies truly benefiting from AI are expanding while traditional software firms contract, with the economy showing bubble-like concentration in AI hardware and energy sectors.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  8. \n

    Halliburton enhances seismic workflow creation with Amazon Bedrock and Generative AI<\/strong>\uff08AWS ML Blog\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Halliburton \u4e0e AWS \u751f\u6210\u5f0f AI \u521b\u65b0\u4e2d\u5fc3\u5408\u4f5c\uff0c\u57fa\u4e8e Amazon Bedrock\u3001Nova\u3001Bedrock Knowledge Bases \u548c DynamoDB \u6784\u5efa\u4e86\u9762\u5411\u5730\u9707\u6570\u636e\u5904\u7406\u5f15\u64ce\u7684 AI \u52a9\u624b\u3002\u8be5\u65b9\u6848\u5141\u8bb8\u5730\u7403\u79d1\u5b66\u5bb6\u548c\u6570\u636e\u79d1\u5b66\u5bb6\u901a\u8fc7\u81ea\u7136\u8bed\u8a00\u4ea4\u4e92\u914d\u7f6e\u539f\u672c\u9700\u8981\u624b\u52a8\u8bbe\u7f6e\u7ea6 100 \u4e2a\u4e13\u4e1a\u5de5\u5177\u7684\u590d\u6742\u5730\u9707\u5de5\u4f5c\u6d41\u3002\u8bc4\u4f30\u7ed3\u679c\u663e\u793a\uff0c\u5de5\u4f5c\u6d41\u521b\u5efa\u6548\u7387\u63d0\u5347\u9ad8\u8fbe 95%\uff0c\u663e\u8457\u964d\u4f4e\u4e86\u4e13\u4e1a\u95e8\u69db\u548c\u64cd\u4f5c\u9519\u8bef\u7387\uff0c\u4e3a\u80fd\u6e90\u52d8\u63a2\u9886\u57df\u7684\u751f\u6210\u5f0f AI \u5e94\u7528\u63d0\u4f9b\u4e86\u53ef\u590d\u7528\u7684\u6280\u672f\u8303\u5f0f\u3002<\/p>\n

    English Summary:<\/strong> Halliburton partnered with the AWS Generative AI Innovation Center to build an AI assistant for their Seismic Engine using Amazon Bedrock, Nova, Bedrock Knowledge Bases, and DynamoDB. The solution enables geoscientists and data scientists to configure complex seismic workflows through natural language interaction instead of manually configuring approximately 100 specialized tools.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  9. \n

    Running Codex safely at OpenAI<\/strong>\uff08OpenAI News\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>OpenAI \u53d1\u5e03\u6280\u672f\u535a\u5ba2\uff0c\u8be6\u7ec6\u4ecb\u7ecd\u4e86\u5185\u90e8\u5982\u4f55\u5b89\u5168\u90e8\u7f72 Codex \u7f16\u7801\u667a\u80fd\u4f53\u3002\u6838\u5fc3\u63aa\u65bd\u5305\u62ec\uff1a\u901a\u8fc7\u6c99\u7bb1\u548c\u5ba1\u6279\u7b56\u7565\u63a7\u5236\u6267\u884c\u8fb9\u754c\uff0c\u4f4e\u98ce\u9669\u7684\u65e5\u5e38\u64cd\u4f5c\u81ea\u52a8\u901a\u8fc7\uff0c\u9ad8\u98ce\u9669\u64cd\u4f5c\u9700\u4eba\u5de5\u5ba1\u6838\uff1b\u5b9e\u65bd\u53d7\u7ba1\u7f51\u7edc\u7b56\u7565\u9650\u5236\u51fa\u7ad9\u8bbf\u95ee\uff1b\u4f7f\u7528\u89c4\u5219\u533a\u5206\u547d\u4ee4\u5b89\u5168\u7b49\u7ea7\uff1b\u901a\u8fc7 macOS \u6258\u7ba1\u504f\u597d\u8bbe\u7f6e\u548c\u672c\u5730\u9700\u6c42\u6587\u4ef6\u5b9e\u73b0\u914d\u7f6e\u7ba1\u7406\uff1b\u5bfc\u51fa OpenTelemetry \u65e5\u5fd7\u5b9e\u73b0\u667a\u80fd\u4f53\u539f\u751f\u7684\u53ef\u89c2\u6d4b\u6027\u5ba1\u8ba1\u3002\u8fd9\u4e9b\u5b9e\u8df5\u4e3a\u4f01\u4e1a\u5b89\u5168\u91c7\u7528 AI \u7f16\u7801\u52a9\u624b\u63d0\u4f9b\u4e86\u53ef\u843d\u5730\u7684\u53c2\u8003\u6846\u67b6\u3002<\/p>\n

    English Summary:<\/strong> OpenAI published a technical blog detailing how they safely deploy the Codex coding agent internally. Key measures include sandboxing and approval policies to control execution boundaries, with low-risk actions proceeding automatically while high-risk actions require human review; managed network policies restricting outbound access; rule-based command safety classification; configuration management via macOS managed preferences and local requirements files; and OpenTelemetry log export for agent-native observability and auditing.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  10. \n

    [AINews] GPT-Realtime-2, -Translate, and -Whisper: new SOTA realtime voice APIs<\/strong>\uff08Latent Space\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>OpenAI \u53d1\u5e03\u4e09\u6b3e\u5b9e\u65f6\u8bed\u97f3 API \u65b0\u6a21\u578b\uff1aGPT-Realtime-2\u3001GPT-Realtime-Translate \u548c GPT-Realtime-Whisper\u3002GPT-Realtime-2 \u88ab\u5b9a\u4f4d\u4e3a OpenAI \u8fc4\u4eca\u6700\u667a\u80fd\u7684\u8bed\u97f3\u6a21\u578b\uff0c\u5728 Big Bench Audio \u57fa\u51c6\u4e0a\u8f83 realtime-1.5 \u63d0\u5347 15.2%\uff0c\u652f\u6301 GPT-5 \u7ea7\u522b\u7684\u63a8\u7406\u3001\u5e76\u884c\u5de5\u5177\u8c03\u7528\u300132K\u2192128K \u4e0a\u4e0b\u6587\u6269\u5c55\u53ca\u4e94\u7ea7\u53ef\u8c03\u63a8\u7406\u5f3a\u5ea6\u3002GPT-Realtime-Translate \u652f\u6301 70 \u591a\u79cd\u8bed\u8a00\u5b9e\u65f6\u4e92\u8bd1\uff0cGPT-Realtime-Whisper \u63d0\u4f9b\u6d41\u5f0f\u8f6c\u5f55\u3002\u65b0\u529f\u80fd\u8fd8\u5305\u62ec\u9884\u54cd\u5e94\u77ed\u8bed\u3001\u66f4\u5f3a\u7684\u9886\u57df\u672f\u8bed\u7406\u89e3\u548c\u53ef\u63a7\u7684\u8bed\u6c14\u8868\u8fbe\u3002<\/p>\n

    English Summary:<\/strong> OpenAI released three new real-time voice API models: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. GPT-Realtime-2 is positioned as OpenAI's most intelligent voice model yet, achieving a 15.2% improvement over realtime-1.5 on Big Bench Audio, supporting GPT-5-class reasoning, parallel tool calls, 32K\u2192128K context expansion, and five adjustable reasoning levels. GPT-Realtime-Translate supports real-time translation across 70+ languages, while GPT-Realtime-Whisper provides streaming transcription.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  11. \n

    Improving token efficiency in GitHub Agentic Workflows<\/strong>\uff08GitHub AI\/ML\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>GitHub \u5de5\u7a0b\u56e2\u961f\u5206\u4eab\u4e86\u5728\u751f\u4ea7\u73af\u5883\u4e2d\u4f18\u5316 Agentic Workflows \u7684\u5b9e\u8df5\u7ecf\u9a8c\u3002\u901a\u8fc7\u5728 API \u4ee3\u7406\u5c42\u7edf\u4e00\u91c7\u96c6 token \u4f7f\u7528\u6570\u636e\uff0c\u56e2\u961f\u53d1\u73b0\u672a\u4f7f\u7528\u7684 MCP \u5de5\u5177\u662f\u4e3b\u8981\u5f00\u9500\u6765\u6e90\u2014\u201440 \u4e2a\u5de5\u5177\u7684 JSON Schema \u53ef\u80fd\u4e3a\u6bcf\u6b21\u8bf7\u6c42\u589e\u52a0 10-15KB \u4e0a\u4e0b\u6587\u3002\u4ed6\u4eec\u6784\u5efa\u4e86\u6bcf\u65e5 Token \u5ba1\u8ba1\u4e0e\u4f18\u5316\u5de5\u4f5c\u6d41\uff0c\u5c06\u6570\u636e\u83b7\u53d6\u64cd\u4f5c\u4ece LLM \u63a8\u7406\u5faa\u73af\u4e2d\u5265\u79bb\uff0c\u6539\u7528 GitHub CLI \u9884\u4e0b\u8f7d\u6216\u4ee3\u7406\u65b9\u5f0f\u6267\u884c\u3002\u4f18\u5316\u540e\uff0cAuto-Triage Issues \u5de5\u4f5c\u6d41\u8282\u7701 62% \u7684 Effective Tokens\uff0cSecurity Guard \u548c Smoke Claude \u5206\u522b\u8282\u7701 43% \u548c 59%\u3002\u6587\u7ae0\u8fd8\u63d0\u51fa ET\uff08Effective Tokens\uff09\u6307\u6807\u6765\u6807\u51c6\u5316\u4e0d\u540c\u6a21\u578b\u95f4\u7684\u6210\u672c\u6bd4\u8f83\uff0c\u5e76\u5f3a\u8c03\u8fd0\u884c\u9891\u7387\u4e0e\u5355\u6b21\u8282\u7701\u540c\u7b49\u91cd\u8981\u3002<\/p>\n

    English Summary:<\/strong> GitHub's engineering team shares practical experience optimizing Agentic Workflows in production. By instrumenting token usage at the API proxy layer, they identified unused MCP tools as a major cost driver\u201440 tools' JSON schemas could add 10-15KB per request. They built daily token auditing and optimization workflows, moving data-fetching operations out of the LLM reasoning loop using GitHub CLI pre-downloads or proxy approaches. Results show Auto-Triage Issues reduced Effective Tokens by 62%, while Security Guard and Smoke Claude achieved 43% and 59% savings respectively. The article introduces the ET (Effective Tokens) metric for normalizing costs across models and emphasizes that run frequency matters as much as per-run savings.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  12. \n

    Agent pull requests are everywhere. Here\u2019s how to review them.<\/strong>\uff08GitHub AI\/ML\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>GitHub \u53d1\u5e03\u4e86\u4e00\u4efd\u5ba1\u67e5 AI \u751f\u6210 Pull Request \u7684\u5b9e\u7528\u6307\u5357\u3002\u968f\u7740 Copilot Code Review \u5904\u7406\u91cf\u7a81\u7834 6000 \u4e07\u6b21\u4e14\u589e\u957f 10 \u500d\uff0c\u8d85\u8fc7\u4e94\u5206\u4e4b\u4e00\u7684\u4ee3\u7801\u5ba1\u67e5\u6d89\u53ca AI \u4ee3\u7406\u3002\u6587\u7ae0\u6307\u51fa\uff0c\u7814\u7a76\u8868\u660e AI \u751f\u6210\u7684\u4ee3\u7801\u6bd4\u4eba\u5de5\u4ee3\u7801\u5f15\u5165\u66f4\u591a\u5197\u4f59\u548c\u6280\u672f\u503a\u52a1\u3002\u5ba1\u67e5\u8005\u5e94\u91cd\u70b9\u5173\u6ce8\u4e94\u5927\u7ea2\u65d7\uff1aCI \u5f31\u5316\uff08\u5982\u5220\u9664\u6d4b\u8bd5\u3001\u964d\u4f4e\u8986\u76d6\u7387\u9608\u503c\uff09\u3001\u4ee3\u7801\u590d\u7528\u76f2\u533a\uff08\u91cd\u590d\u9020\u8f6e\u5b50\uff09\u3001\u5e7b\u89c9\u6b63\u786e\u6027\uff08\u7f16\u8bd1\u901a\u8fc7\u4f46\u903b\u8f91\u9519\u8bef\uff09\u3001\u4ee3\u7406\u5931\u8054\uff08\u5927\u578b PR \u7f3a\u4e4f\u5b9e\u65bd\u8ba1\u5212\uff09\u4ee5\u53ca\u5de5\u4f5c\u6d41\u4e2d\u7684\u4e0d\u53ef\u4fe1\u8f93\u5165\uff08\u63d0\u793a\u6ce8\u5165\u98ce\u9669\uff09\u3002\u5efa\u8bae\u5148\u8ba9 Copilot \u8fdb\u884c\u81ea\u52a8\u5316\u5ba1\u67e5\u5904\u7406\u673a\u68b0\u6027\u95ee\u9898\uff0c\u4eba\u5de5\u4e13\u6ce8\u4e8e\u5173\u952e\u8def\u5f84\u8ffd\u8e2a\u548c\u5224\u65ad\uff0c\u5e76\u63d0\u4f9b\u4e86 10 \u5206\u949f\u5feb\u901f\u5ba1\u67e5\u6e05\u5355\u3002<\/p>\n

    English Summary:<\/strong> GitHub published a practical guide for reviewing AI-generated pull requests. As Copilot Code Review processed over 60 million reviews with 10x growth, more than one in five code reviews now involve AI agents. The article notes research showing AI-generated code introduces more redundancy and technical debt than human-written code. Reviewers should watch for five red flags: CI gaming (removing tests, lowering coverage), code reuse blindness (reinventing utilities), hallucinated correctness (compiles but wrong), agentic ghosting (large PRs without implementation plans), and untrusted input in workflows (prompt injection risks). The guide recommends letting Copilot handle mechanical checks first while humans focus on tracing critical paths and judgment, providing a 10-minute review checklist.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  13. \n

    Notes from inside China's AI labs<\/strong>\uff08Interconnects\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Interconnects AI \u535a\u4e3b Nathan Lambert \u5206\u4eab\u4e86\u8d70\u8bbf\u4e2d\u56fd\u4e3b\u8981 AI \u5b9e\u9a8c\u5ba4\u7684\u89c2\u5bdf\u3002\u4ed6\u53d1\u73b0\u4e2d\u56fd\u7814\u7a76\u8005\u5c55\u73b0\u51fa\u6781\u81f4\u7684\u8c26\u900a\u4e0e\u52a1\u5b9e\uff1a\u5b66\u751f\u76f4\u63a5\u53c2\u4e0e\u6838\u5fc3\u6a21\u578b\u5f00\u53d1\u3001\u8f83\u5c11\u53d7\u4e2a\u4eba\u58f0\u8a89\u535a\u5f08\u5e72\u6270\u3001\u4e13\u6ce8\u4e8e\u5de5\u7a0b\u5b9e\u73b0\u800c\u975e\u54f2\u5b66\u8ba8\u8bba\u3002\u4e2d\u56fd AI \u751f\u6001\u66f4\u50cf\u534f\u4f5c\u7f51\u7edc\u800c\u975e\u5bf9\u6297\u90e8\u843d\uff0c\u5404\u5b9e\u9a8c\u5ba4\u5bf9 DeepSeek \u7684\u7814\u7a76\u54c1\u5473\u548c ByteDance \u7684\u5e02\u573a\u8868\u73b0\u4e92\u8868\u656c\u610f\u3002\u4e0e\u897f\u65b9\u4e0d\u540c\uff0c\u4e2d\u56fd\u79d1\u6280\u516c\u53f8\u666e\u904d\u91c7\u7528"\u81ea\u5efa\u800c\u975e\u8d2d\u4e70"\u7b56\u7565\u2014\u2014\u7f8e\u56e2\u3001\u5c0f\u7c73\u7b49\u975e\u4f20\u7edf AI \u516c\u53f8\u90fd\u5728\u8bad\u7ec3\u901a\u7528\u5927\u6a21\u578b\u4ee5\u638c\u63a7\u6280\u672f\u6808\u3002\u6570\u636e\u4ea7\u4e1a\u76f8\u5bf9\u6b20\u53d1\u8fbe\uff0cRL \u73af\u5883\u591a\u7531\u7814\u7a76\u8005\u81ea\u884c\u6784\u5efa\u3002\u5c3d\u7ba1\u53d7\u9650\u4e8e Nvidia \u82af\u7247\u4f9b\u5e94\uff0c\u534e\u4e3a\u82af\u7247\u5728\u63a8\u7406\u7aef\u83b7\u5f97\u79ef\u6781\u8bc4\u4ef7\u3002\u6587\u7ae0\u6700\u540e\u8868\u8fbe\u4e86\u5bf9\u5168\u7403 AI \u793e\u533a\u5206\u88c2\u98ce\u9669\u7684\u62c5\u5fe7\uff0c\u547c\u5401\u5f00\u653e\u751f\u6001\u5171\u540c\u53d1\u5c55\u3002<\/p>\n

    English Summary:<\/strong> Interconnects AI author Nathan Lambert shares observations from visiting leading Chinese AI labs. He found Chinese researchers display extreme humility and pragmatism: students directly participate in core model development, with less ego-driven politics and more focus on engineering over philosophical debates. China's AI ecosystem resembles a collaborative network rather than competing tribes, with mutual respect for DeepSeek's research taste and ByteDance's market dominance. Unlike the West, Chinese tech companies widely adopt a 'build-not-buy' mentality\u2014non-traditional AI firms like Meituan and Xiaomi train general-purpose LLMs to control their stack. The data industry is less developed, with RL environments often built in-house. Despite Nvidia chip constraints, Huawei chips receive positive remarks for inference. The article concludes with concerns about global AI community fragmentation and calls for open ecosystem collaboration.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  14. \n

    Scaling Trusted Access for Cyber with GPT-5.5 and GPT-5.5-Cyber<\/strong>\uff08OpenAI News\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>OpenAI \u6269\u5c55\u4e86 Trusted Access for Cyber\uff08TAC\uff09\u8ba1\u5212\uff0c\u63a8\u51fa GPT-5.5-Cyber \u9650\u91cf\u9884\u89c8\u7248\uff0c\u4e13\u4e3a\u7f51\u7edc\u9632\u5fa1\u8005\u8bbe\u8ba1\u3002TAC \u662f\u57fa\u4e8e\u8eab\u4efd\u548c\u4fe1\u4efb\u9a8c\u8bc1\u7684\u6846\u67b6\uff0c\u901a\u8fc7\u9a8c\u8bc1\u7684\u9632\u5fa1\u8005\u53ef\u83b7\u5f97\u66f4\u4f4e\u7684\u5206\u7c7b\u5668\u62d2\u7edd\u7387\uff0c\u652f\u6301\u6f0f\u6d1e\u8bc6\u522b\u3001\u6076\u610f\u8f6f\u4ef6\u5206\u6790\u3001\u4e8c\u8fdb\u5236\u9006\u5411\u5de5\u7a0b\u3001\u68c0\u6d4b\u5de5\u7a0b\u7b49\u5de5\u4f5c\u6d41\uff0c\u540c\u65f6\u7ee7\u7eed\u963b\u6b62\u51ed\u8bc1\u7a83\u53d6\u3001\u6076\u610f\u8f6f\u4ef6\u90e8\u7f72\u7b49\u6076\u610f\u6d3b\u52a8\u3002GPT-5.5-Cyber \u6bd4\u6807\u51c6\u7248\u66f4\u5bbd\u677e\uff0c\u652f\u6301\u6388\u6743\u7ea2\u961f\u6d4b\u8bd5\u548c\u6e17\u900f\u6d4b\u8bd5\u7b49\u4e13\u4e1a\u5de5\u4f5c\u6d41\uff0c\u4f46\u9700\u66f4\u5f3a\u7684\u9a8c\u8bc1\u548c\u8d26\u6237\u7ea7\u63a7\u5236\u3002OpenAI \u4e0e Cisco\u3001CrowdStrike\u3001Palo Alto Networks \u7b49\u5b89\u5168\u5382\u5546\u5408\u4f5c\uff0c\u6784\u5efa\u4ece\u6f0f\u6d1e\u4fee\u8865\u5230\u7f51\u7edc\u9632\u62a4\u7684\u5b89\u5168\u98de\u8f6e\u3002\u4e2a\u4eba\u7528\u6237\u53ef\u5728 chatgpt.com\/cyber \u7533\u8bf7\u9a8c\u8bc1\uff0c\u4f01\u4e1a\u7528\u6237\u53ef\u8054\u7cfb OpenAI \u4ee3\u8868\u3002<\/p>\n

    English Summary:<\/strong> OpenAI expanded its Trusted Access for Cyber (TAC) program with a limited preview of GPT-5.5-Cyber, designed specifically for cyber defenders. TAC is an identity and trust-based framework where verified defenders receive lower classifier refusals for authorized defensive workflows including vulnerability identification, malware analysis, binary reverse engineering, and detection engineering, while malicious activities like credential theft remain blocked. GPT-5.5-Cyber is more permissive than the standard version, supporting specialized workflows like authorized red teaming and penetration testing with stronger verification and account-level controls. OpenAI partners with security vendors including Cisco, CrowdStrike, and Palo Alto Networks to build a security flywheel from vulnerability patching to network protection. Individual users can apply at chatgpt.com\/cyber; enterprises can contact OpenAI representatives.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n

  15. \n

    Ollama is now powered by MLX on Apple Silicon in preview<\/strong>\uff08Ollama Blog\uff09<\/p>\n

    \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Ollama \u53d1\u5e03\u9884\u89c8\u7248\uff0c\u5728 Apple Silicon \u4e0a\u91c7\u7528 Apple \u7684\u673a\u5668\u5b66\u4e60\u6846\u67b6 MLX\uff0c\u5b9e\u73b0\u6027\u80fd\u5927\u5e45\u63d0\u5347\u3002\u5728 M5 \u7cfb\u5217\u82af\u7247\u4e0a\uff0cOllama \u5229\u7528\u65b0\u7684 GPU Neural Accelerators \u52a0\u901f\u9996 token \u65f6\u95f4\uff08TTFT\uff09\u548c\u751f\u6210\u901f\u5ea6\u3002\u6d4b\u8bd5\u663e\u793a\uff0c\u4f7f\u7528 NVFP4 \u91cf\u5316\u7684 Qwen3.5-35B-A3B \u6a21\u578b\uff0c\u9884\u586b\u5145\u901f\u5ea6\u8fbe 1851 token\/s\uff0c\u89e3\u7801\u901f\u5ea6\u8fbe 134 token\/s\u3002\u65b0\u7248\u672c\u5f15\u5165 NVFP4 \u652f\u6301\uff0c\u5728\u4fdd\u6301\u6a21\u578b\u7cbe\u5ea6\u7684\u540c\u65f6\u964d\u4f4e\u5185\u5b58\u5e26\u5bbd\u548c\u5b58\u50a8\u9700\u6c42\uff0c\u4e0e\u751f\u4ea7\u73af\u5883\u63a8\u7406\u7ed3\u679c\u4e00\u81f4\u3002\u7f13\u5b58\u7cfb\u7edf\u4e5f\u5f97\u5230\u5347\u7ea7\uff1a\u8de8\u5bf9\u8bdd\u590d\u7528\u7f13\u5b58\u964d\u4f4e\u5185\u5b58\u5360\u7528\u3001\u667a\u80fd\u68c0\u67e5\u70b9\u51cf\u5c11\u63d0\u793a\u5904\u7406\u3001\u5171\u4eab\u524d\u7f00\u5728\u5206\u652f\u573a\u666f\u4e0b\u5b58\u6d3b\u66f4\u4e45\u3002\u7528\u6237\u9700 32GB \u4ee5\u4e0a\u7edf\u4e00\u5185\u5b58\u7684 Mac\uff0c\u53ef\u901a\u8fc7 Ollama 0.19 \u542f\u52a8 Claude Code \u6216 OpenClaw \u7b49\u7f16\u7801\u4ee3\u7406\u3002<\/p>\n

    English Summary:<\/strong> Ollama released a preview version powered by Apple's machine learning framework MLX on Apple Silicon, delivering significant performance improvements. On M5 series chips, Ollama leverages new GPU Neural Accelerators to accelerate both time-to-first-token (TTFT) and generation speed. Testing shows the Qwen3.5-35B-A3B model with NVFP4 quantization achieves 1851 tokens\/s prefill and 134 tokens\/s decode speeds. The new version introduces NVFP4 support, maintaining model accuracy while reducing memory bandwidth and storage requirements, matching production inference results. The caching system is also upgraded: cross-conversation cache reuse reduces memory footprint, intelligent checkpoints minimize prompt processing, and shared prefixes survive longer in branching scenarios. Users need a Mac with 32GB+ unified memory and can launch coding agents like Claude Code or OpenClaw via Ollama 0.19.<\/p>\n

    \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"

    \u65e5\u671f\uff1a2026-05-11 \u672c\u671f\u805a\u7126\uff1a\u91cd\u70b9\u5173\u6ce8\u6a21\u578b\u53d1\u5e03\u4e0e release notes\u3001\u5b98\u65b9 engineeri […]<\/p>\n","protected":false},"author":0,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[7],"tags":[],"class_list":["post-393","post","type-post","status-publish","format-standard","hentry","category-ai-daily"],"_links":{"self":[{"href":"http:\/\/www.faiyi.com\/index.php?rest_route=\/wp\/v2\/posts\/393","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.faiyi.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.faiyi.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"http:\/\/www.faiyi.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=393"}],"version-history":[{"count":0,"href":"http:\/\/www.faiyi.com\/index.php?rest_route=\/wp\/v2\/posts\/393\/revisions"}],"wp:attachment":[{"href":"http:\/\/www.faiyi.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=393"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.faiyi.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=393"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.faiyi.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=393"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}