AI Daily · Nov 15
Nov 14 06:00 – Nov 15 06:00 (Beijing time) · 15 items
Updated 11/15 06:00 (Beijing time)
Today's highlights
- CursorCursor discloses security incident investigation: malicious Claude Code extension affected up to 5,000 usersCursor published the findings of its security incident investigation, stating that malicious and unauthorized copies of the Anthropic Claude Code extension affected up to 5,000 users, and that it has contacted each affected customer individually; among users who did not use privacy mode, fewer than 0.05% had coding agent trace data leaked.
- AnthropicClaude Code Projects update: more proactive cross-thread coordinationAnthropic officially announced a major update to the Projects feature in Claude Code: within a project, Claude can coordinate tasks across parallel threads, and the coordinator is more proactive, answering questions in the project chat and reporting on the progress of each thread.
- Google AIGemini 4 Hydrogen demo goes from idea to finished filmOn its official Gemini X account, Google showcased the video generation capabilities of Gemini 4 Hydrogen: with only custom skills and a simple request, it automatically retrieved public domain archive footage and edited it into a narrated promotional short film about the element hydrogen.
Models and products
- Cursor
Cursor discloses security incident investigation: malicious Claude Code extension affected up to 5,000 users
Cursor published the findings of its security incident investigation, stating that malicious and unauthorized copies of the Anthropic Claude Code extension affected up to 5,000 users, and that it has contacted each affected customer individually; among users who did not use privacy mode, fewer than 0.05% had coding agent trace data leaked.
Anthropic
Claude Code Projects update: more proactive cross-thread coordination
Anthropic officially announced a major update to the Projects feature in Claude Code: within a project, Claude can coordinate tasks across parallel threads, and the coordinator is more proactive, answering questions in the project chat and reporting on the progress of each thread.
Google AI
Gemini 4 Hydrogen demo goes from idea to finished film
On its official Gemini X account, Google showcased the video generation capabilities of Gemini 4 Hydrogen: with only custom skills and a simple request, it automatically retrieved public domain archive footage and edited it into a narrated promotional short film about the element hydrogen.
xAI
xAI integrates more products into unified Grok experience
xAI officially announced that it is integrating more @spacexai products into a unified Grok experience, enabling users to create, build, and work from a single entry point.
xAI
Musk: Grok text watermarking enabled only when legally required or requested by users
Musk said on X that Grok's text watermarking is enabled only when legally required or upon users' active request. This responds to external concerns about the labeling of AI-generated content.
- GitHub
GitHub releases 1-minute overview of Octoverse report
GitHub officially releases a 1-minute overview of the Octoverse report, covering the number of pull requests initiated by AI agents on GitHub, how teams are adjusting their workflows with GitHub Copilot, and the most popular programming languages on GitHub.
Anthropic
Claude Code desktop version supports custom themes
Anthropic announced that the Claude Code desktop version allows custom themes via mods, supporting color scheme changes and animated backgrounds. Users can simply tell Claude the look they want, or use the theme-examples plugin to get started quickly.
xAI
xAI says it has improved Grok @Bot efficiency
Musk said on X that xAI has improved the efficiency of Grok @Bot, without disclosing specific details of the changes.
OpenCode
opencode rewrites OpenCode Browse in Rust
opencode officially states that it rewrote OpenCode Browse in Rust and reimplemented the required playwrite/patchwrite parts, spending only $60.
Media reports
- Related companies
Google AIThird-party report
Google executive runs Win11 on iPhone Duo using Gemini
Amar Rehi, head of product and design at Google AI Studio, used the Gemini 4 Carbon model to develop the WindowsiOS project, running Windows 11 on Arm 24H2 via a virtual machine on Apple's first foldable phone, the iPhone Duo, with touch support; the project is open-sourced on GitHub.
- Related companies
DeepSeekThird-party report
DeepSeek frontend code shows "quota exhausted" prompt, possibly signaling consumer paid plans
Developers found quota management logic and clues about a "Coding Plan" package in the frontend code of chat.deepseek.com, speculating that DeepSeek will launch consumer paid plans for the first time; the official pricing, tiers, and launch date have not yet been announced.
- Related companies
Google AIThird-party report
Google validates complex mathematical proofs with multi-agent Teamwork system
Google has extended its Antigravity multi-agent system, using Gemini-driven specialized agents to form Teamwork, completing a formal proof of the Weil conjecture with about 28,800 sub-agent calls generating about 1.8 million lines of Lean code, and reproducing a related formal proof for the 3D Navier-Stokes equations.
- Related companies
OpenAIThird-party report
OpenAI restructures embodied AI team, plans to make GPT-6 Astra the robot "brain"
OpenAI confirmed the restructuring of its embodied AI team, merging robotics researchers into the core research team and moving the applied robotics and hardware teams into the applied division, and said that general models including GPT-6 Astra can already be used directly for some robotics tasks.
- Related companies
MetaThird-party report
Meta's Muse launches call-making feature
Meta's AI agent Muse has launched a call-making feature, available to its highest-tier paid subscribers at a subscription price starting at $19.99 per month, allowing it to contact businesses, insurance companies, and others on users' behalf. Meta says the feature does not always complete tasks successfully.
- Related companies
OpenAIThird-party report
Former OpenAI board member who fired Sam Altman discusses trust in AI CEOs
Helen Toner, executive director of the Center for Security and Emerging Technology at Georgetown University, said in an interview that we still cannot trust the CEOs of AI companies. She previously served on the OpenAI board and took part in firing Sam Altman.
Generates a long image you can save and share.