GraphAHA: Graph-Based Adaptive Search for Test-Time Code Generation
This research presents a new approach to improving code generation efficiency, which is relevant for developers and researchers working with large language models.
DailyVOL 2026.09.15197 items
Apple integrates Claude and GPT-5.6 into Siri AI, while Microsoft releases humanist AI guidelines
Singapore·Tuesday, 15 September 2026·DAILY · 08:00 SGT
2 Sep 2026. Apple has integrated Anthropic's Claude and OpenAI's GPT-5.6 into its Siri AI framework, allowing users to select these third-party models through the 'Ask...' menu. Microsoft released a 37-page humanist AI behavior guideline emphasizing human control over AI systems. Shanghai AI Lab's Intern physical world model W0 for force-tactile robotics and Yushu Robotics' upgraded G1+ humanoid robot with six enhancements highlight AI's expanding applications in robotics. Analog Devices' $135M acquisition of Alif Semiconductor supports its physical intelligence strategy to move AI processing from cloud to edge devices. GraphAHA addresses test-time code generation challenges with graph-based adaptive search, while RoofLang provides a domain-specific language for optimizing LLM inference systems.
01
Models
3 itemsShanghai AI Lab ships Intern physical world model W0 for force-tactile robotics
This is a new AI model from a Chinese lab focused on force-tactile robotics, which could be relevant for those working in related fields or interested in the latest developments in physical world AI models.
Chinese content苹果代码显示：Siri AI 可接入 Claude 或 GPT-5.6
朋友发现苹果代码里，Siri AI 能用 Claude 或 GPT-5.6 替换，未来可能让 Siri 更好用。
IT之家Source
02
Products
3 itemsPortable Computer available on Windows PCs with NVIDIA RTX GPUs
This allows users to run the AI tool locally on their Windows PC with an RTX GPU, avoiding cloud dependency for some tasks.
Yushu Robotics releases upgraded humanoid robot G1+
The G1+ is a more affordable and capable humanoid robot for research and development, offering significant improvements in motion and perception compared to its predecessor.
IT之家Source
03
Industry
3 itemsDoing AI in big tech super apps is surprisingly hard
For users of apps like Meituan or WeChat, this explains why their AI features might be slow to develop or change, and why some features may be removed or replaced.
Chinese contentAnalog Devices花费1.35亿美元收购Alif Semiconductor，押注“物理智能”
朋友A告诉朋友B，Analog Devices花了1.35亿美元收购Alif，是为了把AI直接用在设备上，而不是在云端算，这和之前的方式不太一样。
Chinese content微软发布 37 页人文主义 AI 行为准则：人比 AI 更重要
微软这个大公司发了份很长的准则，说人比 AI 重要，模型不能失控，和之前 OpenAI 的事件有关。
IT之家Source
04
Papers
3 itemsRoofLang: Enabling AI-Driven Architecting of LLM Inference Systems
This matters because it introduces a new tool, RoofLang, that could significantly improve the performance of LLMs like DeepSeek V4-series by optimizing their inference architecture, potentially leading to faster processing and better user experience.
Debiasing LLM-as-a-Judge evaluation with TraceJudgeBench
This paper provides a detailed analysis of how debiasing methods affect the accuracy of LLMs used as judges, which is important for anyone working with RAG or agent-workflow evaluation.
New method improves efficiency of subquadratic attention models
This research provides a concrete method to improve the energy efficiency of large language models, which is important for deploying them on edge devices.
05
How-to
3 itemsGenerative AI customization spectrum on AWS
For developers looking to implement generative AI, AWS's 8-step framework offers a clear path from simple prompt engineering to building custom models, helping them choose the right approach for their needs.
LottieFiles opens motion design skills for UI animation
For developers working on UI design, this provides a structured framework to inject professional motion design judgment into AI models, helping create more effective and polished user interfaces.
197Stories
85First-party
53New models
69Sources
AITOP · generated by the editorial system