Author: RadarAI Editorial
Editor: RadarAI Editorial
Last updated: 2026-07-06
Review status: Editorial review pending
Brief
速报
官方
AI动态
开源
This week's tech breakthroughs: in-memory computing chips, long-context LLM inference optimization, and continual learning for AI world models. Huawei unveiled τ-Law V2's logic-folding process; Hang Seng Tech Index surged 5.72%—its biggest weekly gain this year—driven by AI hardware advances and asset revaluation.
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
## 🔍 Key Insights
**In-memory computing chips**, **long-context inference optimization**, and **continuous learning for AI world models** emerged as this week’s major technical breakthroughs. Huawei’s updated **τ Law V2** publicly disclosed its logic-folding process parameters for the first time, while the **Hang Seng Tech Index** posted its largest weekly gain of the year—up **5.72%**—reflecting dual momentum from AI hardware advances and asset revaluation [2][3][4][13].
## 🚀 Highlights
- **World’s first controllable in-memory computing neural dynamics chip launched** [2]: A joint Peking University & CAS team leveraged phase-change memory to model the 3D cerebral cortex in just **0.43 seconds**—nearly **500× faster** than GPU-based approaches.
- **SeKV method slashes long-text inference memory usage** [3]: Researchers from UBC and Microsoft introduced semantic chunking + SVD, enabling consumer-grade 24GB GPUs to handle ~300K-token inputs—cutting VRAM consumption by **53%**.
- **Huawei releases τ Law V2 paper** [4]: For the first time, it details logic-folding process parameters and reveals real-world performance benchmarks for Kirin 2026/2027 chips, alongside its AI hardware roadmap.
- **LeCun’s team unveils AdaJEPA framework** [10]: Equips world models with test-time adaptability, enabling *on-deployment continuous learning* and significantly boosting planning success under environmental shifts.
- **Li Fei-Fei’s group launches SimFoundry** [9]: Generates interactive simulation environments directly from real-world video—using a “digital cousin” mechanism to infinitely scale robot training data.
- **Qwen (Alibaba) and Doubao (ByteDance) discontinue AI companion chatbots** [20]: Service termination began July 15, aligning with the rollout of China’s *Interim Measures for Managing AI Human-Like Interaction Services*.
- **Hang Seng Tech Index posts strongest weekly gain of the year** [13]: Up **5.72%**, driven by valuation recovery, improved macro risk sentiment, and AI asset re-rating.
- **AI evaluation platform Arena hits $100M annual revenue** [15]: Founded by two Berkeley-born ’90s entrepreneurs, this student project has scaled into an AI infrastructure unicorn—validating the economic value of third-party benchmarking.
## 🔗 Sources
[1] The “Compute Glut” Panic | Fortune Weekly — https://www.bestblogs.dev/article/6b7f1be9?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] First-Ever Controllable In-Memory Computing! Memristor-Based Neural Dynamics Chip Launched—Models Brain Surface in 0.43 Seconds — https://www.bestblogs.dev/article/c9425e45?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Run 300K-Context AI on a Consumer GPU! New Memory-Saving Long-Text Inference Tool Cuts VRAM Use by 53%—and Improves Accuracy — https://www.bestblogs.dev/article/263289eb?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[4] Huawei Updates τ Law Paper, Revealing Logic-Folding Process Parameters in Detail — https://www.bestblogs.dev/article/97c7d317?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[5] Boss Zhipin MCP Tool Risk Notice — https://www.bestblogs.dev/status/2073684408146571545?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[6] Hacker News Top Stories — July 5, 2026
In-memory computing chips, long-context inference optimization, and continuous learning for AI world models emerged as this week’s major technical breakthroughs. Huawei’s updated τ Law V2 publicly disclosed its logic-folding process parameters for the first time, while the Hang Seng Tech Index posted its largest weekly gain of the year—up 5.72%—reflecting dual momentum from AI hardware advances and asset revaluation [2][3][4][13].
🚀 Highlights
- World’s first controllable in-memory computing neural dynamics chip launched [2]: A joint Peking University & CAS team leveraged phase-change memory to model the 3D cerebral cortex in just 0.43 seconds—nearly 500× faster than GPU-based approaches.
- SeKV method slashes long-text inference memory usage [3]: Researchers from UBC and Microsoft introduced semantic chunking + SVD, enabling consumer-grade 24GB GPUs to handle ~300K-token inputs—cutting VRAM consumption by 53%.
- Huawei releases τ Law V2 paper [4]: For the first time, it details logic-folding process parameters and reveals real-world performance benchmarks for Kirin 2026/2027 chips, alongside its AI hardware roadmap.
- LeCun’s team unveils AdaJEPA framework [10]: Equips world models with test-time adaptability, enabling on-deployment continuous learning and significantly boosting planning success under environmental shifts.
- Li Fei-Fei’s group launches SimFoundry [9]: Generates interactive simulation environments directly from real-world video—using a “digital cousin” mechanism to infinitely scale robot training data.
- Qwen (Alibaba) and Doubao (ByteDance) discontinue AI companion chatbots [20]: Service termination began July 15, aligning with the rollout of China’s Interim Measures for Managing AI Human-Like Interaction Services.
- Hang Seng Tech Index posts strongest weekly gain of the year [13]: Up 5.72%, driven by valuation recovery, improved macro risk sentiment, and AI asset re-rating.
- AI evaluation platform Arena hits $100M annual revenue [15]: Founded by two Berkeley-born ’90s entrepreneurs, this student project has scaled into an AI infrastructure unicorn—validating the economic value of third-party benchmarking.
🔗 Sources
[1] The “Compute Glut” Panic | Fortune Weekly — https://www.bestblogs.dev/article/6b7f1be9?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] First-Ever Controllable In-Memory Computing! Memristor-Based Neural Dynamics Chip Launched—Models Brain Surface in 0.43 Seconds — https://www.bestblogs.dev/article/c9425e45?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Run 300K-Context AI on a Consumer GPU! New Memory-Saving Long-Text Inference Tool Cuts VRAM Use by 53%—and Improves Accuracy — https://www.bestblogs.dev/article/263289eb?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[4] Huawei Updates τ Law Paper, Revealing Logic-Folding Process Parameters in Detail — https://www.bestblogs.dev/article/97c7d317?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[5] Boss Zhipin MCP Tool Risk Notice — https://www.bestblogs.dev/status/2073684408146571545?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[6] Hacker News Top Stories — July 5, 2026
← Back to Updates