Daily AI News

Google's WikiSkill gives AI agents a persistent memory of past mistakes to sharpen future performance

Google Research unveiled WikiSkill, a framework that equips AI agents with a persistent wiki‑like knowledge base, allowing them to record both successes and failures across runs. This memory enables agents to refine future behavior, especially larger models that can leverage accumulated insights. The approach aims to improve agent reliability and learning efficiency.

Read source →

Building Custom Batched Ensemble Weather Forecasting with NVIDIA Earth2Studio

A tutorial demonstrates how to construct a custom batched ensemble weather forecasting pipeline using NVIDIA’s Earth2Studio. By integrating the FCN prognostic model with real‑time atmospheric data from GFS, users can run large‑scale ensemble predictions within a Colab environment. This showcases Earth2Studio’s flexibility for scientific AI applications.

Read source →

Meet ‘Code-as-World’: An Agentic Loop That Rewrites Real Videos Into Executable MuJoCo Physics Programs

The Code-as-World system translates real‑world video footage into editable MuJoCo physics programs, creating executable scene representations. These generated worlds are then validated and used to train physical‑reasoning models, bridging perception and simulation. The method offers a new avenue for grounding AI in real‑world dynamics.

Read source →

OpenAI cuts off Cursor after SpaceX acquisition, citing Musk's history of breaking contracts

OpenAI has terminated its integration with the AI coding assistant Cursor following Cursor’s acquisition by SpaceX, citing concerns over Elon Musk’s contract history. Cursor’s co‑founder downplayed the impact, noting that OpenAI models contributed only a small fraction of the tool’s traffic. The split highlights tensions in AI tool partnerships.

Read source →

Sony Music, Warner sue Anthropic, alleging a “brazen campaign” of intellectual property theft

Sony Music and Warner Music have filed a lawsuit against Anthropic, accusing the firm of a widespread campaign to steal intellectual property by generating copyrighted content. The complaint alleges illegal piracy and misuse of protected works in Anthropic’s models. The case could set a precedent for AI‑generated content liability.

Read source →

AI-generated videos are already displacing actors and livestreamers across China's entertainment industry

In China, AI‑generated short dramas now comprise 95% of the 128,000 releases in Q1 2026, leading to actors and livestreamers losing jobs to synthetic avatars. The rapid adoption has sparked labor disputes and concerns over the ethical use of AI in entertainment. The trend underscores the disruptive impact of generative video technology.

Read source →

“We’re not doing 30 bets a year”: Vijay Pande on betting small after running $4 billion at a16z

Vijay Pande discusses his shift from managing a $4 billion biotech fund at a16z to leading the smaller AI‑native venture VZVC, emphasizing a strategy of making fewer, more focused bets. He argues that biology is transitioning toward an engineering discipline, requiring open data and shared resources. The interview reflects evolving investment philosophies in AI‑driven biotech.

Read source →

Musicians-turned-detectives are hunting for AI grifters

A community of musicians‑turned‑detectives is actively identifying and exposing AI‑generated music that mimics human artists, aiming to protect creators from plagiarism and fraud. As generative audio tools improve, the market is flooded with synthetic tracks that can mislead listeners and monetize stolen styles. Their efforts raise awareness of authenticity in the digital music landscape.

Read source →