\u65e5\u671f\uff1a2026-05-04<\/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
Artificial Analysis \u6700\u65b0\u6a21\u578b\u6392\u540d\u89c2\u5bdf<\/strong>\uff08Artificial Analysis\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Artificial Analysis \u53d1\u5e03\u6700\u65b0\u6a21\u578b\u7efc\u5408\u6392\u540d\uff0cGPT-5.5 (xhigh) \u4ee5 60 \u5206\u4f4d\u5c45 Intelligence Index \u699c\u9996\uff0cClaude Opus 4.7 (max) \u4e0e Gemini 3.1 Pro Preview \u5e76\u5217\u7b2c\u4e09\u3002\u8f93\u51fa\u901f\u5ea6\u65b9\u9762\uff0cMercury 2 \u4ee5 778 tokens\/s \u9886\u5148\uff1b\u5ef6\u8fdf\u6700\u4f4e\u7684\u662f Ministral 3 3B\uff080.47 \u79d2\uff09\u3002\u5f00\u6e90\u6a21\u578b\u4e2d\uff0cKimi K2.6 \u6392\u540d\u6700\u9ad8\uff0854 \u5206\uff09\u3002\u5e73\u53f0\u65b0\u589e Intelligence Index v4.0\uff0c\u6db5\u76d6 GDPval-AA\u3001Terminal-Bench Hard\u3001Humanity's Last Exam \u7b49 10 \u9879\u8bc4\u6d4b\uff0c\u5e76\u7ee7\u7eed\u63d0\u4f9b\u4ef7\u683c\u3001\u4e0a\u4e0b\u6587\u7a97\u53e3\u3001\u5f00\u653e\u6027\u7b49\u591a\u7ef4\u5ea6\u5bf9\u6bd4\u5de5\u5177\u3002<\/p>\n English Summary:<\/strong> Artificial Analysis released its latest model rankings: GPT-5.5 (xhigh) leads the Intelligence Index at 60, followed by GPT-5.5 (high) at 59, with Claude Opus 4.7 (max) and Gemini 3.1 Pro Preview tied at 57. Mercury 2 tops output speed at 778 tokens\/s, while Ministral 3 3B has the lowest latency at 0.47s. Among open-weights models, Kimi K2.6 ranks highest at 54. The platform now uses Intelligence Index v4.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\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\u65f6\u4efb\u52a1\u4e2d\u8868\u73b0\u66f4\u4e3a\u4e25\u8c28\u4e00\u81f4\u3002\u65b0\u6a21\u578b\u652f\u6301\u66f4\u9ad8\u5206\u8fa8\u7387\u56fe\u50cf\u8f93\u5165\uff08\u957f\u8fb9\u6700\u9ad8 2576 \u50cf\u7d20\uff09\uff0c\u89c6\u89c9\u80fd\u529b\u5927\u5e45\u589e\u5f3a\u3002\u65b0\u589e xhigh effort \u6863\u4f4d\uff0cClaude Code \u9ed8\u8ba4 effort \u63d0\u5347\u81f3 xhigh\u3002API \u5b9a\u4ef7\u7ef4\u6301\u4e0d\u53d8\uff08\u8f93\u5165 $5\/\u767e\u4e07 tokens\uff0c\u8f93\u51fa $25\/\u767e\u4e07 tokens\uff09\u3002\u6a21\u578b\u5df2\u90e8\u7f72\u81ea\u52a8\u7f51\u7edc\u5b89\u5168\u9632\u62a4\u673a\u5236\uff0c\u5e76\u63a8\u51fa Cyber Verification Program \u4f9b\u5b89\u5168\u7814\u7a76\u4eba\u5458\u7533\u8bf7\u5408\u6cd5\u4f7f\u7528\u3002Cursor\u3001Replit\u3001Vercel \u7b49\u5408\u4f5c\u4f19\u4f34\u53cd\u9988\u663e\u793a\uff0c\u4ee3\u7801\u8d28\u91cf\u3001\u5de5\u5177\u8c03\u7528\u51c6\u786e\u7387\u548c\u957f\u7a0b\u81ea\u4e3b\u6027\u5747\u6709\u660e\u663e\u6539\u5584\u3002<\/p>\n English Summary:<\/strong> Anthropic announced Claude Opus 4.7, featuring notable improvements in advanced software engineering over 4.6, with stronger performance on complex, long-running tasks. The model now supports higher-resolution images up to 2,576 pixels on the long edge. A new xhigh effort level is introduced, with Claude Code defaulting to xhigh for Opus 4.7. Pricing remains unchanged at $5\/M input and $25\/M output tokens. The release includes automated cyber safeguards and a Cyber Verification Program for security researchers. Early partners including Cursor, Replit, and Vercel reported significant gains in code quality, tool accuracy, and long-horizon autonomy.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\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\u4e0b\u964d\u4e8b\u4ef6\u7684\u590d\u76d8\u62a5\u544a\uff0c\u786e\u8ba4\u4e09\u5904\u72ec\u7acb\u53d8\u66f4\u5bfc\u81f4\u7528\u6237\u4f53\u9a8c\u95ee\u9898\uff1a3 \u6708 4 \u65e5\u5c06\u9ed8\u8ba4 effort \u4ece high \u964d\u81f3 medium\uff084 \u6708 7 \u65e5\u56de\u6eda\uff09\uff1b3 \u6708 26 \u65e5\u7f13\u5b58\u4f18\u5316 bug \u5bfc\u81f4\u4f1a\u8bdd\u95f2\u7f6e\u8d85 1 \u5c0f\u65f6\u540e\u6301\u7eed\u4e22\u5931\u63a8\u7406\u5386\u53f2\uff084 \u6708 10 \u65e5\u4fee\u590d\uff09\uff1b4 \u6708 16 \u65e5\u7cfb\u7edf\u63d0\u793a\u8bcd\u65b0\u589e\u5b57\u6570\u9650\u5236\u6307\u4ee4\u610f\u5916\u964d\u4f4e\u7f16\u7801\u8d28\u91cf\uff084 \u6708 20 \u65e5\u56de\u6eda\uff09\u3002\u56e2\u961f\u5df2\u5411\u6240\u6709\u8ba2\u9605\u8005\u91cd\u7f6e\u4f7f\u7528\u989d\u5ea6\uff0c\u5e76\u627f\u8bfa\u52a0\u5f3a\u5185\u90e8\u6d4b\u8bd5\u6d41\u7a0b\uff0c\u5305\u62ec\u6269\u5927\u5458\u5de5\u4f7f\u7528\u516c\u5171\u7248\u672c\u8303\u56f4\u3001\u5b8c\u5584 Code Review \u5de5\u5177\uff0c\u4ee5\u53ca\u9488\u5bf9\u7cfb\u7edf\u63d0\u793a\u8bcd\u53d8\u66f4\u5efa\u7acb\u66f4\u4e25\u683c\u7684\u8bc4\u4f30\u4e0e\u6e10\u8fdb\u53d1\u5e03\u673a\u5236\u3002<\/p>\n English Summary:<\/strong> Anthropic Engineering published a postmortem on recent Claude Code quality issues, tracing reports to three separate changes: a March 4 default effort reduction from high to medium (reverted April 7); a March 26 caching optimization bug that continuously dropped reasoning history after sessions idle for over an hour (fixed April 10); and an April 16 system prompt change adding length limits that degraded coding quality (reverted April 20). The team reset usage limits for all subscribers and committed to process improvements, including broader internal dogfooding of public builds, enhanced Code Review tooling, and stricter evaluation with gradual rollouts for system prompt changes.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\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\u53d1\u5e03 Managed Agents \u67b6\u6784\u8bbe\u8ba1\u6587\u7ae0\uff0c\u9610\u8ff0\u5982\u4f55\u901a\u8fc7\u89e3\u8026"\u5927\u8111"\uff08Claude \u53ca\u5176 harness\uff09\u4e0e"\u53cc\u624b"\uff08sandbox \u4e0e\u5de5\u5177\uff09\u4ee5\u53ca"\u4f1a\u8bdd"\uff08\u4e8b\u4ef6\u65e5\u5fd7\uff09\u6765\u6784\u5efa\u53ef\u6269\u5c55\u7684\u957f\u671f\u8fd0\u884c Agent \u6258\u7ba1\u670d\u52a1\u3002\u6838\u5fc3\u8bbe\u8ba1\u501f\u9274\u64cd\u4f5c\u7cfb\u7edf\u865a\u62df\u5316\u601d\u60f3\uff0c\u901a\u8fc7\u6807\u51c6\u5316\u63a5\u53e3\uff08execute\u3001provision\u3001emitEvent\u3001getSession \u7b49\uff09\u4f7f\u5404\u7ec4\u4ef6\u53ef\u72ec\u7acb\u5931\u8d25\u3001\u66ff\u6362\u4e0e\u6269\u5c55\u3002\u89e3\u8026\u540e p50 \u9996 token \u5ef6\u8fdf\u964d\u4f4e\u7ea6 60%\uff0cp95 \u964d\u4f4e\u8d85 90%\uff1b\u540c\u65f6\u652f\u6301\u591a brain \u4e0e\u591a hand \u67b6\u6784\uff0c\u5141\u8bb8\u8de8 VPC \u8c03\u7528\u8d44\u6e90\u800c\u65e0\u9700\u7f51\u7edc\u5bf9\u7b49\u8fde\u63a5\u3002\u5b89\u5168\u5c42\u9762\uff0c\u51ed\u8bc1\u5b58\u50a8\u4e8e vault \u5916\u90e8\uff0c\u901a\u8fc7 MCP \u4ee3\u7406\u8c03\u7528\uff0c\u786e\u4fdd sandbox \u5185\u4ee3\u7801\u65e0\u6cd5\u63a5\u89e6\u654f\u611f\u4ee4\u724c\u3002<\/p>\n English Summary:<\/strong> Anthropic's Engineering Blog published a deep dive on Managed Agents architecture, explaining how decoupling the brain (Claude and its harness) from the hands (sandboxes and tools) and the session (event log) enables scalable, long-running agent hosting. Drawing from OS virtualization principles, standardized interfaces like execute, provision, emitEvent, and getSession allow components to fail and scale independently. Decoupling reduced p50 time-to-first-token by roughly 60% and p95 by over 90%. The architecture supports many brains and many hands, enabling cross-VPC resource access without network peering. Security is enforced by storing credentials in an external vault and routing MCP tool calls through a proxy, ensuring generated code in sandboxes cannot access sensitive tokens.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n \u2018This is fine\u2019 creator says AI startup stole his art<\/strong>\uff08TechCrunch AI\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>\u77e5\u540d\u7f51\u7edc\u6f2b\u753b\u300aThis is fine\u300b\u4f5c\u8005 KC Green \u6307\u63a7 AI \u521b\u4e1a\u516c\u53f8 Artisan \u672a\u7ecf\u6388\u6743\u5728\u5176\u5730\u94c1\u5e7f\u544a\u4e2d\u4f7f\u7528\u8be5 meme \u5f62\u8c61\u3002\u5e7f\u544a\u5c06\u539f\u6f2b\u753b\u4e2d\u7684\u53f0\u8bcd\u6539\u4e3a my pipeline is on fire\uff0c\u5e76\u914d\u4ee5 Hire Ava the AI BDR \u7684\u6587\u6848\u3002Green \u8868\u793a\u672a\u540c\u610f\u8be5\u4f7f\u7528\u65b9\u5f0f\uff0c\u5e76\u79f0\u4f5c\u54c1\u88ab\u50cf AI \u4e00\u6837\u7a83\u53d6\uff0c\u547c\u5401\u89c1\u5230\u5e7f\u544a\u7684\u4eba\u8fdb\u884c\u6d82\u9e26\u7834\u574f\u3002Artisan \u56de\u5e94\u79f0\u5c0a\u91cd Green \u53ca\u5176\u4f5c\u54c1\uff0c\u5df2\u4e3b\u52a8\u8054\u7cfb\u5e76\u5b89\u6392\u6c9f\u901a\u3002\u8be5\u516c\u53f8\u6b64\u524d\u66fe\u56e0 stop hiring humans \u7cfb\u5217\u5e7f\u544a\u724c\u5f15\u53d1\u4e89\u8bae\u3002Green \u8868\u793a\u5c06\u5bfb\u6c42\u6cd5\u5f8b\u4ee3\u8868\u7ef4\u6743\uff0c\u540c\u65f6\u611f\u53f9\u4e0d\u5f97\u4e0d\u4e3a\u6b64\u8017\u8d39\u672c\u5e94\u7528\u4e8e\u521b\u4f5c\u7684\u65f6\u95f4\u4e0e\u7cbe\u529b\u3002<\/p>\n English Summary:<\/strong> KC Green, creator of the iconic This is fine meme, accused AI startup Artisan of using his artwork without permission in a subway ad campaign. The ad adapted the comic with the line my pipeline is on fire and text urging viewers to Hire Ava the AI BDR. Green stated he did not agree to the usage and called it stolen like AI steals, asking people to vandalize the ads if seen. Artisan responded with respect for Green and said they reached out to schedule a conversation. The company previously drew controversy with billboards urging businesses to stop hiring humans. Green told TechCrunch he will seek legal representation, expressing frustration at having to divert time from creating comics to navigate the legal system.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n Cloudflare Builds High-Performance Infrastructure for Running LLMs<\/strong>\uff08InfoQ AI\/ML\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Cloudflare \u5ba3\u5e03\u5728\u5176\u5168\u7403\u7f51\u7edc\u90e8\u7f72\u4e13\u4e3a\u5927\u89c4\u6a21\u8bed\u8a00\u6a21\u578b\u8bbe\u8ba1\u7684\u9ad8\u6027\u80fd\u57fa\u7840\u8bbe\u65bd\u3002\u4e3a\u89e3\u51b3 LLM \u5bf9\u6602\u8d35\u786c\u4ef6\u7684\u4f9d\u8d56\u53ca\u9ad8\u541e\u5410\u6587\u672c\u5904\u7406\u9700\u6c42\uff0cCloudflare \u91c7\u7528\u300c\u5206\u79bb\u5f0f\u9884\u586b\u5145\uff08disaggregated prefill\uff09\u300d\u67b6\u6784\uff0c\u5c06\u6a21\u578b\u5904\u7406\u62c6\u5206\u4e3a\u4e24\u4e2a\u9636\u6bb5\uff1a\u9884\u586b\u5145\u9636\u6bb5\uff08\u5904\u7406\u8f93\u5165 token\uff0c\u8ba1\u7b97\u5bc6\u96c6\uff09\u548c\u89e3\u7801\u9636\u6bb5\uff08\u751f\u6210\u8f93\u51fa token\uff0c\u5185\u5b58\u5bc6\u96c6\uff09\uff0c\u5206\u522b\u7531\u4e0d\u540c\u673a\u5668\u5904\u7406\u3002\u540c\u65f6\u63a8\u51fa\u81ea\u7814 AI \u63a8\u7406\u5f15\u64ce Infire\uff0c\u53ef\u5728\u591a GPU \u95f4\u9ad8\u6548\u8fd0\u884c\u5927\u6a21\u578b\uff0c\u51cf\u5c11\u5185\u5b58\u5360\u7528\u5e76\u52a0\u5feb\u542f\u52a8\u901f\u5ea6\u3002\u4ee5 Kimi K2.5\uff08\u8d85 1 \u4e07\u4ebf\u53c2\u6570\u3001\u7ea6 560GB\uff09\u4e3a\u4f8b\uff0c\u8be5\u4f18\u5316\u663e\u8457\u63d0\u5347\u4e86\u8d85\u5927\u6a21\u578b\u7684\u54cd\u5e94\u901f\u5ea6\u548c\u8fd0\u884c\u6548\u7387\u3002<\/p>\n English Summary:<\/strong> Cloudflare announced new infrastructure designed to run large language models across its global network. To address the costly hardware requirements and high-throughput text processing demands of LLMs, Cloudflare employs a "disaggregated prefill" architecture that splits model processing into two stages: prefill (input token processing, compute-bound) and decode (output generation, memory-bound), handled by separate machines. The company also introduced Infire, a custom AI inference engine that runs large models across multiple GPUs more efficiently, reduces memory usage, and enables faster model startup. For models like Kimi K2.5 (over 1 trillion parameters, ~560GB), these optimizations significantly improve response times and operational efficiency.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n [AINews] AI Engineer World's Fair \u2014 Autoresearch, Memory, World Models, Tokenmaxxing, Agentic Commerce, and Vertical AI Call for Speakers<\/strong>\uff08Latent Space\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Latent Space \u53d1\u5e03 AI Engineer World's Fair \u7b2c\u4e8c\u6ce2\u6f14\u8bb2\u8005\u62db\u52df\u901a\u77e5\u3002\u8be5\u4f1a\u8bae\u805a\u7126 AI \u5de5\u7a0b\u524d\u6cbf\u8bae\u9898\uff0c\u5305\u62ec\u81ea\u52a8\u7814\u7a76\uff08Autoresearch\uff09\u3001\u8bb0\u5fc6\u7cfb\u7edf\uff08Memory\uff09\u3001\u4e16\u754c\u6a21\u578b\uff08World Models\uff09\u3001Token \u4f18\u5316\uff08Tokenmaxxing\uff09\u3001\u667a\u80fd\u4f53\u5546\u4e1a\uff08Agentic Commerce\uff09\u53ca\u5782\u76f4\u9886\u57df AI\uff08Vertical AI\uff09\u7b49\u65b9\u5411\u3002\u540c\u65f6\u6587\u7ae0\u56de\u987e\u4e86\u8fd1\u671f AI \u9886\u57df\u52a8\u6001\uff1a\u7f8e\u56fd\u56fd\u9632\u90e8\u4e0e\u4e03\u5bb6\u524d\u6cbf AI \u53ca\u57fa\u7840\u8bbe\u65bd\u516c\u53f8\u8fbe\u6210\u5408\u4f5c\uff0c\u5c06 AI \u80fd\u529b\u90e8\u7f72\u81f3\u6d89\u5bc6\u7f51\u7edc\uff1bOpenAI CEO Sam Altman \u5f3a\u8c03\u300c\u6784\u5efa\u589e\u5f3a\u4eba\u7c7b\u7684\u5de5\u5177\uff0c\u800c\u975e\u53d6\u4ee3\u4eba\u7c7b\u7684\u5b9e\u4f53\u300d\uff1bCodex \u4ea7\u54c1\u5728\u5f00\u53d1\u8005\u4e2d\u83b7\u5f97\u79ef\u6781\u53cd\u54cd\uff1bARC Prize \u8bc4\u6d4b\u663e\u793a GPT-5.5 \u548c Opus 4.7 \u5728 ARC-AGI-3 \u57fa\u51c6\u4e0a\u8868\u73b0\u6709\u9650\uff0c\u5f15\u53d1\u5bf9\u6a21\u578b\u80fd\u529b\u7684\u8ba8\u8bba\u3002<\/p>\n English Summary:<\/strong> Latent Space announced the Wave 2 Call for Speakers for the AI Engineer World's Fair, focusing on frontier AI engineering topics including Autoresearch, Memory systems, World Models, Tokenmaxxing, Agentic Commerce, and Vertical AI. The article also recaps recent AI developments: the U.S. Department of Defense announced partnerships with seven frontier AI and infrastructure companies for classified network deployments; OpenAI CEO Sam Altman emphasized building tools to augment rather than replace humans; Codex received positive developer adoption feedback; and ARC Prize benchmarks showed GPT-5.5 at 0.43% and Opus 4.7 at 0.18% on ARC-AGI-3, sparking discussions on model capabilities.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n AWS Transform now automates BI migration to Amazon Quick in days<\/strong>\uff08AWS ML Blog\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>AWS \u5ba3\u5e03 AWS Transform \u73b0\u5df2\u652f\u6301\u5728\u6570\u5929\u5185\u81ea\u52a8\u5b8c\u6210 BI\uff08\u5546\u4e1a\u667a\u80fd\uff09\u8fc1\u79fb\u81f3 Amazon QuickSight\u3002\u8be5\u65b9\u6848\u57fa\u4e8e Amazon Bedrock \u63d0\u4f9b\u5e95\u5c42 AI \u80fd\u529b\uff0c\u901a\u8fc7 Amazon Bedrock AgentCore \u4f5c\u4e3a\u5b89\u5168\u8fd0\u884c\u65f6\u73af\u5883\uff0c\u5b9e\u73b0\u51ed\u8bc1\u7ba1\u7406\u548c IAM \u8bbf\u95ee\u63a7\u5236\u3002AWS Transform \u4f5c\u4e3a\u534f\u4f5c\u5f0f\u4f01\u4e1a IT \u8f6c\u578b\u5de5\u4f5c\u53f0\uff0c\u63d0\u4f9b\u57fa\u4e8e\u5bf9\u8bdd\u7684\u8fc1\u79fb\u4f5c\u4e1a\u521b\u5efa\u4e0e\u7ba1\u7406\u754c\u9762\uff1b\u5408\u4f5c\u4f19\u4f34 Wavicle \u7684 BI \u8fc1\u79fb\u7ecf\u9a8c\u88ab\u7f16\u7801\u4e3a\u667a\u80fd\u4f53\u903b\u8f91\u3002\u6574\u4e2a\u6d41\u7a0b\u5728\u5ba2\u6237\u81ea\u6709 AWS \u8d26\u6237\u5185\u8fd0\u884c\uff0c\u6570\u636e\u65e0\u9700\u79bb\u5f00\u73af\u5883\uff0c\u6d88\u9664\u4e86\u4f20\u7edf\u8fc1\u79fb\u9879\u76ee\u4e2d\u7684\u5b89\u5168\u4e0e\u91c7\u8d2d\u6469\u64e6\u3002\u76ee\u6807\u670d\u52a1 Amazon QuickSight \u63d0\u4f9b\u65e0\u670d\u52a1\u5668\u6269\u5c55\u80fd\u529b\u3001SPICE \u5185\u5b58\u5f15\u64ce\u6027\u80fd\u53ca\u4e0e AWS \u6570\u636e\u670d\u52a1\u7684\u539f\u751f\u96c6\u6210\u3002<\/p>\n English Summary:<\/strong> AWS announced that AWS Transform now automates BI migration to Amazon QuickSight in days. The solution leverages Amazon Bedrock for underlying AI capabilities and Amazon Bedrock AgentCore as a secure runtime environment for credential management and IAM-based access control. AWS Transform serves as a collaborative enterprise IT transformation workbench with a conversational interface for creating and managing migration jobs. Partner Wavicle's BI migration expertise is encoded into agent logic. The entire process runs within the customer's own AWS account with no data leaving the environment, eliminating security and procurement friction typical of migration projects. The target service Amazon QuickSight offers serverless scalability, SPICE in-memory engine performance, and native integration with AWS data services.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n [AINews] Agents for Everything Else: Codex for Knowledge Work, Claude for Creative Work<\/strong>\uff08Latent Space\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>Latent Space \u6587\u7ae0\u63a2\u8ba8\u300c\u7f16\u7801\u667a\u80fd\u4f53\u6b63\u5728\u7a81\u7834\u8fb9\u754c\u300d\u7684\u8d8b\u52bf\uff0c\u6307\u51fa Claude \u548c Codex \u8fd1\u671f\u5747\u6709\u91cd\u5927\u53d1\u5e03\uff0cClaude \u5728\u58f0\u91cf\u4e0a\u6301\u7eed\u9886\u5148\u3002OpenAI \u5c06 Codex \u4ece\u300c\u7f16\u7801\u667a\u80fd\u4f53\u300d\u6218\u7565\u6269\u5c55\u4e3a\u300c\u901a\u7528\u77e5\u8bc6\u5de5\u4f5c\u667a\u80fd\u4f53\u300d\uff0cSam Altman \u7684\u8ddf\u8fdb\u8868\u6001\u6210\u4e3a\u5f53\u65e5\u6700\u53d7\u5173\u6ce8\u7684\u4ea7\u54c1\u52a8\u6001\u4e4b\u4e00\u3002Anthropic \u63a8\u51fa Claude Security\uff0c\u7531 Opus 4.7 \u9a71\u52a8\uff0c\u53ef\u626b\u63cf\u4ee3\u7801\u4ed3\u5e93\u6f0f\u6d1e\u3001\u9a8c\u8bc1\u53d1\u73b0\u5e76\u63d0\u4f9b\u4fee\u590d\u5efa\u8bae\uff1bCursor \u540c\u6b65\u63a8\u51fa Cursor Security Review\uff0c\u652f\u6301\u6301\u7eed PR \u5ba1\u67e5\u548c\u5b9a\u65f6\u4ee3\u7801\u5e93\u626b\u63cf\uff0c\u6807\u5fd7\u7740\u6a21\u578b\u5382\u5546\u6b63\u5f0f\u8fdb\u5165 DevSecOps \u9886\u57df\u3002\u6b64\u5916\uff0cQwen \u53d1\u5e03\u53ef\u89e3\u91ca\u6027\u5de5\u5177\u5957\u4ef6 Qwen-Scope\uff08\u7a00\u758f\u81ea\u7f16\u7801\u5668\uff09\uff0cAnthropic \u53d1\u5e03\u57fa\u4e8e 100 \u4e07\u6b21 Claude \u5bf9\u8bdd\u7684\u5927\u89c4\u6a21\u6307\u5bfc\/\u8c04\u5a9a\u884c\u4e3a\u7814\u7a76\uff0c\u5e76\u5c06\u53d1\u73b0\u76f4\u63a5\u5e94\u7528\u4e8e Opus 4.7 \u548c Mythos Preview \u7684\u8bad\u7ec3\u6539\u8fdb\u3002<\/p>\n English Summary:<\/strong> Latent Space reflects on the trend of "coding agents breaking containment," noting major releases from both Claude and Codex, with Claude continuing to dominate impression counts. OpenAI is strategically expanding Codex from a "coding agent" to a "computer-use agent" for general knowledge work, with Sam Altman's follow-up comments becoming the day's biggest product news. Anthropic launched Claude Security, powered by Opus 4.7, which scans repositories for vulnerabilities, validates findings, and suggests fixes; Cursor shipped Cursor Security Review with always-on PR review and scheduled codebase scans\u2014clear examples of model vendors entering established DevSecOps categories. Additionally, Qwen released Qwen-Scope, an interpretability toolkit with sparse autoencoders, and Anthropic published a large-scale guidance\/sycophancy study based on 1M Claude conversations, directly applying findings to training improvements for Opus 4.7 and Mythos Preview.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n GitHub Copilot CLI for Beginners: Interactive v. non-interactive mode<\/strong>\uff08GitHub AI\/ML\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>GitHub \u53d1\u5e03 Copilot CLI \u521d\u5b66\u8005\u7cfb\u5217\u6587\u7ae0\uff0c\u4ecb\u7ecd\u4ea4\u4e92\u5f0f\uff08interactive\uff09\u4e0e\u975e\u4ea4\u4e92\u5f0f\uff08non-interactive\uff09\u4e24\u79cd\u6a21\u5f0f\u7684\u533a\u522b\u3002\u4ea4\u4e92\u6a21\u5f0f\u901a\u8fc7\u5bf9\u8bdd\u5f0f\u754c\u9762\u5f15\u5bfc\u7528\u6237\u5b8c\u6210\u590d\u6742\u4efb\u52a1\uff0c\u9002\u5408\u9700\u8981\u9010\u6b65\u786e\u8ba4\u548c\u63a2\u7d22\u7684\u573a\u666f\uff1b\u975e\u4ea4\u4e92\u6a21\u5f0f\u5219\u5141\u8bb8\u76f4\u63a5\u6267\u884c\u547d\u4ee4\uff0c\u9002\u7528\u4e8e\u811a\u672c\u5316\u548c\u81ea\u52a8\u5316\u5de5\u4f5c\u6d41\u3002\u8be5\u7cfb\u5217\u4ee5\u89c6\u9891\u548c\u535a\u5ba2\u5f62\u5f0f\u63d0\u4f9b\uff0c\u6db5\u76d6\u4ece\u9996\u4e2a\u63d0\u793a\u8bcd\u5230\u547d\u4ee4\u884c\u9ad8\u6548\u5bfc\u822a\u7684\u5b8c\u6574\u5165\u95e8\u6307\u5357\u3002GitHub \u8fd8\u63d0\u4f9b\u4e86\u76f8\u5173\u8d44\u6e90\uff0c\u5305\u62ec Copilot CLI \u659c\u6760\u547d\u4ee4\u4f7f\u7528\u3001MCP \u670d\u52a1\u5668\u96c6\u6210\u7b49\u8fdb\u9636\u5185\u5bb9\uff0c\u5e2e\u52a9\u5f00\u53d1\u8005\u5145\u5206\u5229\u7528 AI \u8f85\u52a9\u547d\u4ee4\u884c\u5de5\u5177\u63d0\u5347\u751f\u4ea7\u529b\u3002<\/p>\n English Summary:<\/strong> GitHub published a beginner series on Copilot CLI explaining the difference between interactive and non-interactive modes. Interactive mode provides a conversational interface to guide users through complex tasks, suitable for scenarios requiring step-by-step confirmation and exploration; non-interactive mode allows direct command execution, ideal for scripting and automated workflows. The series is available in both video and blog formats, covering everything from first prompts to tips for efficient command line navigation. GitHub also provides additional resources including using Copilot CLI slash commands and integrating MCP servers, helping developers fully leverage AI-assisted command line tools to boost productivity.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n Introducing Advanced Account Security<\/strong>\uff08OpenAI News\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>OpenAI \u63a8\u51fa\u9762\u5411\u9ad8\u98ce\u9669\u7528\u6237\u7684\u9ad8\u7ea7\u8d26\u6237\u5b89\u5168\u529f\u80fd\uff0c\u63d0\u4f9b\u9632\u9493\u9c7c\u767b\u5f55\u3001\u5f3a\u5316\u8d26\u6237\u6062\u590d\u548c\u589e\u5f3a\u4fdd\u62a4\u63aa\u65bd\u3002\u8be5\u529f\u80fd\u8981\u6c42\u4f7f\u7528\u901a\u884c\u5bc6\u94a5\u6216\u7269\u7406\u5b89\u5168\u5bc6\u94a5\u767b\u5f55\uff0c\u7981\u7528\u5bc6\u7801\u548c\u90ae\u4ef6\/SMS \u6062\u590d\u65b9\u5f0f\uff0c\u7f29\u77ed\u4f1a\u8bdd\u6709\u6548\u671f\u5e76\u63d0\u4f9b\u4f1a\u8bdd\u7ba1\u7406\u5de5\u5177\uff0c\u540c\u65f6\u81ea\u52a8\u6392\u9664\u8bad\u7ec3\u6570\u636e\u4f7f\u7528\u3002OpenAI \u4e0e Yubico \u5408\u4f5c\u4e3a\u7528\u6237\u63d0\u4f9b\u5b89\u5168\u5bc6\u94a5\u4f18\u60e0\uff0c\u5e76\u8ba1\u5212\u4ece 2026 \u5e74 6 \u6708\u8d77\u8981\u6c42 Trusted Access for Cyber \u6210\u5458\u5f3a\u5236\u542f\u7528\u8be5\u529f\u80fd\u3002<\/p>\n English Summary:<\/strong> OpenAI introduces Advanced Account Security for high-risk users, offering phishing-resistant login via passkeys or physical security keys, stronger recovery methods, shorter sessions with clearer management, and automatic training exclusion. Partnered with Yubico for discounted security keys. Mandatory for Trusted Access for Cyber members starting June 2026.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n Where the goblins came from<\/strong>\uff08OpenAI News\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>OpenAI \u8be6\u7ec6\u62ab\u9732\u4e86 GPT-5 \u7cfb\u5217\u6a21\u578b\u4e2d\u51fa\u73b0"\u5730\u7cbe\/\u54e5\u5e03\u6797"\u9690\u55bb\u8bcd\u6c47\u5f02\u5e38\u589e\u591a\u73b0\u8c61\u7684\u6839\u6e90\u3002\u8c03\u67e5\u53d1\u73b0\uff0c\u8be5\u95ee\u9898\u6e90\u4e8e\u4e3a"Nerdy"\u4e2a\u6027\u5316\u529f\u80fd\u8bad\u7ec3\u65f6\u8bbe\u7f6e\u7684\u5956\u52b1\u4fe1\u53f7\u65e0\u610f\u4e2d\u504f\u597d\u5305\u542b\u751f\u7269\u9690\u55bb\u7684\u8f93\u51fa\uff0c\u5bfc\u81f4\u8fd9\u4e00\u8bed\u8a00\u98ce\u683c\u4ece\u7279\u5b9a\u4eba\u683c\u8bbe\u7f6e\u6269\u6563\u5230\u6574\u4f53\u6a21\u578b\u884c\u4e3a\u3002\u5c3d\u7ba1 OpenAI \u5df2\u4e8e 3 \u6708\u4e0b\u7ebf\u8be5\u4eba\u683c\u5e76\u4fee\u590d\u8bad\u7ec3\u6570\u636e\uff0c\u4f46 GPT-5.5 \u56e0\u8bad\u7ec3\u65f6\u95f4\u8f83\u65e9\u4ecd\u53d7\u5f71\u54cd\uff0c\u56e2\u961f\u5df2\u901a\u8fc7\u5f00\u53d1\u8005\u63d0\u793a\u8bcd\u7f13\u89e3\u3002\u6b64\u4e8b\u4f8b\u5c55\u793a\u4e86\u5956\u52b1\u4fe1\u53f7\u5982\u4f55\u4ee5\u610f\u60f3\u4e0d\u5230\u7684\u65b9\u5f0f\u5851\u9020\u6a21\u578b\u884c\u4e3a\u3002<\/p>\n English Summary:<\/strong> OpenAI details the root cause of increasing "goblin\/gremlin" metaphors in GPT-5 models. The issue stemmed from reward signals for the "Nerdy" personality feature inadvertently favoring creature metaphors, causing the tic to spread from specific personality settings to general model behavior. The personality was retired in March and training data filtered, though GPT-5.5 still shows effects due to earlier training start. Codex now includes mitigating developer prompts.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n Reading today's open-closed performance gap<\/strong>\uff08Interconnects\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>\u6587\u7ae0\u6df1\u5165\u5206\u6790\u4e86\u5f53\u524d\u5f00\u6e90\u4e0e\u95ed\u6e90\u5927\u6a21\u578b\u4e4b\u95f4\u7684\u6027\u80fd\u5dee\u8ddd\u52a8\u6001\u3002\u4f5c\u8005\u6307\u51fa\uff0c\u5c06\u8fd9\u4e00\u5dee\u8ddd\u7b80\u5316\u4e3a\u5355\u4e00\u6570\u5b57\u4f1a\u63a9\u76d6\u5173\u952e\u7ec6\u8282\uff1a\u8bc4\u6d4b\u57fa\u51c6\u6bcf 12-18 \u4e2a\u6708\u5c31\u4f1a\u968f\u884c\u4e1a\u7126\u70b9\u8f6c\u79fb\u800c\u53d8\u5316\uff0c\u4ece\u65e9\u671f\u7684\u804a\u5929\u3001\u6570\u5b66\u80fd\u529b\u8f6c\u5411\u5f53\u524d\u7684\u590d\u6742\u4ee3\u7801\u548c\u7ec8\u7aef\u4efb\u52a1\u3002\u95ed\u6e90\u524d\u6cbf\u5b9e\u9a8c\u5ba4\u6b63\u6295\u5165\u5de8\u8d44\u638c\u63e1\u73b0\u6709\u7126\u70b9\u9886\u57df\uff0c\u540c\u65f6\u5411\u4f1a\u8ba1\u3001\u6cd5\u5f8b\u3001\u533b\u7597\u7b49\u4e13\u4e1a\u9886\u57df\u6269\u5c55\u3002\u5f00\u6e90\u6a21\u578b\uff08\u5c24\u5176\u662f\u4e2d\u56fd\u5b9e\u9a8c\u5ba4\uff09\u5728\u8ffd\u8d76\u8fc7\u7a0b\u4e2d\u9762\u4e34 RL \u73af\u5883\u6784\u5efa\u548c\u6570\u636e\u83b7\u53d6\u7684\u6311\u6218\uff0c\u4f46\u5728 WeirdML\u3001ARC AGI 2 \u7b49\u5206\u5e03\u5916\u57fa\u51c6\u4e0a\u4ecd\u660e\u663e\u843d\u540e\u3002<\/p>\n English Summary:<\/strong> The article analyzes the evolving open-closed model performance gap, arguing that reducing it to a single number obscures crucial dynamics. Benchmark focus shifts every 12-18 months, moving from chat\/math to complex coding and agentic tasks. Closed frontier labs invest heavily in current domains while pushing into specialized knowledge work. Open models face challenges in RL environment construction and data access, lagging on out-of-distribution benchmarks like WeirdML and ARC AGI 2 despite rapid progress.<\/p>\n \u539f\u6587\u94fe\u63a5<\/a><\/p>\n<\/li>\n Building an emoji list generator with the GitHub Copilot CLI<\/strong>\uff08GitHub AI\/ML\uff09<\/p>\n \u4e2d\u6587\u6458\u8981\uff1a<\/strong>GitHub \u56e2\u961f\u5728 Rubber Duck Thursday \u76f4\u64ad\u4e2d\u4f7f\u7528 GitHub Copilot CLI \u6784\u5efa\u4e86\u4e00\u4e2a\u8868\u60c5\u7b26\u53f7\u5217\u8868\u751f\u6210\u5668\u3002\u8be5\u9879\u76ee\u5229\u7528 Copilot SDK \u5c06\u666e\u901a\u6587\u672c\u5217\u8868\u81ea\u52a8\u8f6c\u6362\u4e3a\u5e26\u76f8\u5173\u8868\u60c5\u7b26\u53f7\u7684\u683c\u5f0f\u5e76\u590d\u5236\u5230\u526a\u8d34\u677f\u3002\u5f00\u53d1\u8fc7\u7a0b\u5c55\u793a\u4e86 Copilot CLI \u7684\u591a\u9879\u529f\u80fd\uff0c\u5305\u62ec Plan \u6a21\u5f0f\u3001Autopilot \u6a21\u5f0f\u3001\u591a\u6a21\u578b\u5de5\u4f5c\u6d41\uff08Claude Sonnet 4.6 \u548c Opus 4.7\uff09\u3001allow-all \u5de5\u5177\u6807\u5fd7\u4ee5\u53ca GitHub MCP \u670d\u52a1\u5668\u3002\u9879\u76ee\u91c7\u7528 OpenTUI \u6784\u5efa\u7ec8\u7aef\u754c\u9762\uff0c\u5df2\u5f00\u6e90\u4f9b\u793e\u533a\u4f7f\u7528\u3002<\/p>\n English Summary:<\/strong> The GitHub team built an emoji list generator using GitHub Copilot CLI during their Rubber Duck Thursday stream. The tool converts plain text lists into emoji-enhanced formats using the Copilot SDK. The development showcased Copilot CLI features including Plan mode, Autopilot mode, multi-model workflows (Claude Sonnet 4.6 and Opus 4.7), the allow-all tools flag, and GitHub MCP server integration. Built with OpenTUI for the terminal interface, the project is open-sourced.<\/p>\n