## 🔍 Core Insights **AI infrastructure** is evolving in depth from the model layer toward **organization-level Agent collaboration** and **on-device efficiency**: **Doubao Large Model 2.1 Pro** and **Feishu 8.0** are pushing Agents to become independent collaboration entities within teams [6][12], while **TypeSafe AI**'s **Jev** model, by abandoning text generation and focusing exclusively on code decision-making, achieves a speedup of **20 to 200 times** [7]. At the same time, the squeeze that the **AI boom** has placed on the **DRAM** supply chain has already passed through to end-device pricing [3], and **Token ROI** and **Harness decomposition** have become focal points for engineering-side reflection [8][10]. ## 🚀 Key Updates - **Doubao Large Model 2.1 Pro version 0915 released** [12]: Comprehensive upgrades in Agent task delivery, multimodal Coding, and inference costs, with image and video Token consumption reduced by over 30%. - **Feishu 8.0 natively integrates with Doubao Work** [6]: Agents enter team workflows as independent collaboration entities, and office AI competition shifts toward organization-level intelligent implementation. - **TypeSafe AI releases System One model Jev** [7]: Founded by a co-inventor of ChatGPT, it abandons text generation and is designed specifically for rapid decision-making within code, running 20 to 200 times faster. - **Ant Lingbo CEO discusses robot deployment** [0]: Advocates training embodied foundation models from scratch and winning through data recipes, judging that the industry will move toward a landscape of a few foundation model companies plus ecosystem collaboration. - **Google opens Claude to all engineers** [1]: Lei Jun visited Unitree and Bilibili on the same day, and HarmonyOS Smart Mobility confirmed that AITO will be led by Seres. - **AI boom triggers DRAM shortage** [3]: Behind the iPhone 18 price increase, the older iPhone 17 bucked the trend and rose by $100 in the US. - **The challenge of measuring Token ROI** [8]: An intern burned 100 million Tokens in a day only to change a dozen lines of code; Input Tokens are the real cost black hole. - **The bitter lesson of Browser Agents** [10]: The stronger the model, the more the Harness should be dismantled, shrinking from manually defined actions to letting the model connect directly to raw CDP interfaces. ## 🔗 Sources [0] Dialogue with Ant Lingbo CEO Zhu Xing: Robots still cannot eat "coarse grains" — https://www.bestblogs.dev/article/f3be36f723?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [1] Morning Brief | Lei Jun visited Unitree and Bilibili on the same day / Luo Yonghao's negative review brought traffic, Wild Man Mr. gained nearly 30,000 followers in a single day / HarmonyOS Smart Mobility confirms AITO cooperation adjustment, Seres takes the lead — https://www.bestblogs.dev/article/945959a3a3?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [3] Tech by the Numbers: Behind the iPhone 18 price increase: How did the AI boom cause a DRAM memory shortage? — https://www.bestblogs.dev/article/48c631c0c9?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [6] Doubao Work plays the "team Agent" card, will WorkBuddy and Qwen follow? — https://www.bestblogs.dev/article/b44044ee95?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [7] TypeSafe AI releases, abandoning text generation