World models and on-device Agent architectures are accelerating toward real-world deployment: Hefei Zhixiang Future raised over RMB 2.1 billion in three months; TuZhan Intelligence open-sourced UniWorld-View—a domestically developed, Ascend-optimized 4D reconstruction model [5]; Tencent proposed that Agents must possess a 'cerebellum' capability—enabling perception-driven closed-loop execution—to ensure tasks are *truly completed*, not merely commands responded to [3].
## 🔍 Core Insights
**World models** and **on-device Agent architectures** are rapidly advancing toward practical deployment: Hefei Zhixiang Future raised over **RMB 2.1 billion** in three months; TuZhan Intelligence open-sourced UniWorld-View—a domestic **4D reconstruction model** optimized for Huawei's Ascend AI chips [5]; Tencent proposed that Agents require a perception-closed-loop '**cerebellum**' capability, emphasizing *task completion* over mere instruction response [3].
## 🚀 Key Updates
- **BaoCut v0.8.2 launches video-frame translation** [0]: Supports Codex Agent–powered automated OCR translation; exported results integrate directly with Jianying (CapCut) for secondary editing.
- **Codex Micro keyboard emerges as a new paradigm for AI Agent control** [1]: Early adopters have already enabled cross-app voice input—highlighting the synergy potential between hardware and Agents.
- **'Digital Dish Developer' and 'Robot Operator' emerge as critical new roles in food service** [2]: Salaries reach up to **RMB 250,000/year**, signaling AI-driven restructuring of talent demand in vertical industries.
- **Tencent Marvis positions itself as an OS-level personal AI assistant** [3]: Built on on-device perception to achieve end-to-end task closure—distinct from conventional conversational Agents.
- **Hefei Zhixiang Future secures over RMB 2.1 billion in funding within three months** [4]: Its native **full-modality architecture** and vision foundation model *vivago R1* propel it to leadership in the multimodal AI space.
- **Domestic world model UniWorld-View tops Li Fei-Fei's team leaderboard** [5]: Enables monocular video-based arbitrary-view synthesis and **4D reconstruction**; fully open-sourced with native Ascend chip support.
- **Peking University open-sources DataFlow-Harness framework** [7]: Embeds data engineering constraints into Code Agents, enabling natural-language generation of **verifiable, reusable DAG pipelines**.
- **WAIC Hall H4 showcases the 'Build in Public' startup paradigm** [6]: AI + social media lowers iteration costs, fostering agile validation—'find users first, then build the product.'
## 🔗 Sources
[0] BaoCut v0.8.2 releases video-frame translation with Agent automation — https://www.bestblogs.dev/status/2080770347977113983?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] First wave of Codex keyboard hands-on reviews: users immediately jailbreak Codex for cross-app voice input — https://www.bestblogs.dev/article/565daade91?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] Salaries up to RMB 250,000/year: two urgently needed new roles emerge! This talent type is now industry-critical → — https://www.bestblogs.dev/article/688c0a190c?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Interview with Tencent VP Songtao Lin: Intent is replacing entry points—Agents need a 'cerebellum' — https://www.bestblogs.dev/article/c79599e995?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[4] Hefei bets again on an AI unicorn: RMB 2.1 billion raised in three months in the multimodal race — https://www.bestblogs.dev/article/fd319f08ca?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[5] Domestic world model tops Li Fei-Fei's team leaderboard! Optimized for domestic hardware
World models and on-device Agent architectures are rapidly advancing toward practical deployment: Hefei Zhixiang Future raised over RMB 2.1 billion in three months; TuZhan Intelligence open-sourced UniWorld-View—a domestic 4D reconstruction model optimized for Huawei's Ascend AI chips [5]; Tencent proposed that Agents require a perception-closed-loop 'cerebellum' capability, emphasizing task completion over mere instruction response [3].
🚀 Key Updates
- BaoCut v0.8.2 launches video-frame translation [0]: Supports Codex Agent–powered automated OCR translation; exported results integrate directly with Jianying (CapCut) for secondary editing.
- Codex Micro keyboard emerges as a new paradigm for AI Agent control [1]: Early adopters have already enabled cross-app voice input—highlighting the synergy potential between hardware and Agents.
- 'Digital Dish Developer' and 'Robot Operator' emerge as critical new roles in food service [2]: Salaries reach up to RMB 250,000/year, signaling AI-driven restructuring of talent demand in vertical industries.
- Tencent Marvis positions itself as an OS-level personal AI assistant [3]: Built on on-device perception to achieve end-to-end task closure—distinct from conventional conversational Agents.
- Hefei Zhixiang Future secures over RMB 2.1 billion in funding within three months [4]: Its native full-modality architecture and vision foundation model vivago R1 propel it to leadership in the multimodal AI space.
- Domestic world model UniWorld-View tops Li Fei-Fei's team leaderboard [5]: Enables monocular video-based arbitrary-view synthesis and 4D reconstruction; fully open-sourced with native Ascend chip support.
- Peking University open-sources DataFlow-Harness framework [7]: Embeds data engineering constraints into Code Agents, enabling natural-language generation of verifiable, reusable DAG pipelines.
- WAIC Hall H4 showcases the 'Build in Public' startup paradigm [6]: AI + social media lowers iteration costs, fostering agile validation—'find users first, then build the product.'
🔗 Sources
[0] BaoCut v0.8.2 releases video-frame translation with Agent automation — https://www.bestblogs.dev/status/2080770347977113983?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] First wave of Codex keyboard hands-on reviews: users immediately jailbreak Codex for cross-app voice input — https://www.bestblogs.dev/article/565daade91?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] Salaries up to RMB 250,000/year: two urgently needed new roles emerge! This talent type is now industry-critical → — https://www.bestblogs.dev/article/688c0a190c?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Interview with Tencent VP Songtao Lin: Intent is replacing entry points—Agents need a 'cerebellum' — https://www.bestblogs.dev/article/c79599e995?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[4] Hefei bets again on an AI unicorn: RMB 2.1 billion raised in three months in the multimodal race — https://www.bestblogs.dev/article/fd319f08ca?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[5] Domestic world model tops Li Fei-Fei's team leaderboard! Optimized for domestic hardware