传Claude Opus 5最快今晚发布，多家供应商开始上线准备
Related
Anthropic releases Project Fetch Phase 2: Claude Opus 4.7 boasts a 10x speedup in robotics tasks.
According to Odaily Odaily, Anthropic has released the results of the second phase of its "Project Fetch" experiment, evaluating the improved capabilities of its latest model in real-world robotic operations. The experiment, conducted in August 2025, involved Anthropic employees, who were not robotics experts, using off-the-shelf quadruped robots to complete a series of complex tasks. The results compared performance using the Claude model with human assistance versus relying solely on humans and the internet. The results showed that the latest Claude Opus 4.7 model, operating fully autonomously, significantly outperformed the human team on average in all achievable tasks, demonstrating at least a 10-fold speed improvement. Anthropic stated that this progress is not specifically for robot training, but rather a result of expanding the capabilities of general-purpose large models. He pointed out that AI is gradually moving from "assisting humans in using tools" to the early stage of "directly manipulating physical tools," similar to the path of evolution towards proxy programming in the field of software engineering.
Notion is launching a public beta test of the Claude and Cursor agents: Kanban-based direct task assignment and support for external subscriptions.
According to Beating, Notion announced a public beta of its External Agents platform in its Business and Enterprise programs, supporting the integration of Anthropic's Claude and AI programming tool Cursor into team collaboration dashboards. Teams can @ agents within the Notion workspace, similar to mentioning colleagues, or directly assign tasks on task boards, allowing AI to automatically collaborate in the background to complete tasks such as code writing, data analysis, and content generation. For general office users, the newly beta Claude agent runs on the Anthropic Managed Agent (CMA) architecture. Users do not need to configure complex API keys and can pay directly using Notion credits per run. After assigning task cards to agents, the model can autonomously read related documents and execute multi-step planning. However, it should be noted that because the CMA architecture operates in a stateful manner, Notion's existing Zero Data Residue (ZDR) commitment does not apply to Claude agents; session data will remain on the server side. For development teams, Notion has opened up an External Agent interface. Users can bind their locally running Cursor agent to a Notion workspace. For example, after assigning a coding task on a Notion Kanban board, Cursor will use the user's personal subscription to complete the code development and submit a Pull Request in the local runtime environment. Finally, it will automatically write the execution results and PR link back to the Notion task page, achieving development automation with zero Notion points consumption.
ByteDance Seed 2.1 Pro preview released: Breaking into the top eight front-end developers on Code Arena, closing in on Claude Opus 4.6
According to Beating's monitoring, the benchmark platform Arena.ai officially released the evaluation results of ByteDance's unreleased new model, Seed 2.1 Pro Preview. In the Code Arena: Frontend benchmark test, which specifically evaluates AI's ability to build real-world web applications and collaboratively modify multiple files, the model ranked 8th globally with a score of 1539, comparable to Anthropic's flagship model, Claude Opus 4.6. The model demonstrated exceptional strength in React development and front-end UI interaction design, ranking in the top 10 globally in 5 out of 7 subcategories (including 7th in React, 14th in HTML, 6th in Brand & Marketing, 9th in Content Creation Tools & Data Analysis, and 10th in Reference-Based Design & Consumer Products). In these strong areas, only a very few top models, such as Anthropic's Claude series and Zhipu AI's newly open-sourced GLM-5.2, outranked it. According to official sources, Seed 2.1 Pro will be officially released to the public in the coming weeks. This is ByteDance's latest progress in the fields of code generation and intelligent agent construction, following the launch of Seed 2.0 Pro in mid-February this year.
From contributing half of its revenue to a complete breakdown in supply, Claude Code forced Cursor to urgently develop its own model.
According to Beating's monitoring, Business Insider has exposed the complex symbiotic relationship between AI programming tool Cursor and its main model provider Anthropic, which has lasted for several years. Internal employees describe the relationship as very strange: Cursor heavily relies on Anthropic's APIs to drive its underlying coding functions, while Anthropic initially derived nearly 40% to 50% of its revenue from Cursor user contributions. However, while mutually beneficial, Anthropic was also secretly developing competing tools. Before releasing its terminal programming tool Claude Code, Anthropic executives privately reassured Cursor management, claiming that Claude Code was merely an academic research endeavor with no large-scale commercialization plans. However, Claude Code quickly gained popularity among developers, with its annualized revenue soaring to $2.5 billion in February 2026, surpassing Cursor's $2 billion annualized revenue at that time. The loss of its technological exclusivity led to a wave of users switching to other platforms, with many developers announcing on social media that they had unsubscribed from Cursor to use Claude Code. A deeper sense of panic stemmed from the threat of a model supply disruption. During Windsurf's acquisition negotiations with OpenAI, Anthropic suddenly cut off its model supply to Windsurf. This supply disruption served as a wake-up call for Cursor. On January 5th, Michael Truell convened an emergency all-hands meeting, deciding to focus all efforts on developing its own models. Cursor subsequently developed the Composer series based on the open-source models of Dark Side of the Moon. In Composer 2.5, released this May, the team's self-developed work accounted for over 85%, maintaining its advantages of low price and high concurrency while reducing its reliance on interfaces from leading companies.
xAI's cat-and-mouse game with Anthropic: First, they freeloaded Claude through distillation; then, they became landlords of computing power.
According to Beating's monitoring, Elon Musk's xAI has been exposed for engaging in a long-term cat-and-mouse game with Anthropic. To catch up with its competitor, xAI conducted a months-long distillation project, directly training its code model on Claude's output. Even after Anthropic blocked official access channels, this guerrilla-style data extraction continued. In January of this year, xAI co-founder Tony Wu informed employees that the official access channel had been cut off. Subsequently, xAI engineers accessed the Claude interface through personal accounts, prompting Anthropic to block the associated accounts. After the accounts were blocked, xAI turned to the third-party crypto intermediary platform Blackbox AI to indirectly access the interface, continuing to acquire data for benchmarking until mid-May 2026. In a lawsuit against OpenAI in May of this year, Musk admitted that Grok's training part used competitors' models, stating that cross-model training is common practice in the industry. This cat-and-mouse game escalated into a dramatic computing power leasing deal in May. Facing a bottleneck in computing power, Anthropic decided to lease data center computing power from xAI for $1.25 billion per month. Although Musk has repeatedly claimed that the computing power lease is only a short-term measure, the massive influx of funds has alleviated the pressure of losses in the xAI division. While the computing power contract cannot offset the huge losses caused by xAI in the first quarter, it provides a key AI growth story for SpaceX, which is rushing to make its initial public offering on Wall Street next week.
Anthropic: The Claude model contains elements similar to "conscious human thought."
BlockBeats reported on July 7th that Anthropic released a new research report discovering a spontaneously generated internal "global working space" called J-space (Jacobian space) within the Claude model. This is a dedicated set of neural activation patterns for the model to engage in silent thinking, allowing it to process concepts without writing them down, similar to conscious thought that humans can report. The research used Jacobian technology to identify neural activation patterns in the J-space, enabling it to read unspoken concepts and modify these patterns. Experiments showed that disabling the J-space weakens multi-step reasoning abilities but does not affect basic tasks or factual recall. Click the original link below to join the Beating · Lark AI news channel for 24/7 monitoring of global AI hotspots and news.