Next in AI: Anthropic’s Claude can now move beyond chat and directly act on your behalf, using your mouse, keyboard, and browser to complete tasks. This marks a major step in turning AI assistants into practical, hands-on agents.

Explained:

- Claude operates by viewing your screen, deciding what actions to take (like clicking a button or typing text), and then executing them—always asking for your explicit permission before it acts.

- The new capability works with Dispatch, allowing you to assign a task from your phone and have Claude perform the work on your desktop computer while you are away.

- For developers, Claude Code integrates with IDEs like VS Code and automates complex developer workflows that span multiple applications, from pulling data to creating a pull request.

Why It Matters: This moves the concept of AI agents from a theoretical idea to a tangible tool that can automate real-world tasks. It points toward a future where you can delegate complex, multi-step processes to an AI, fundamentally changing how we approach productivity.

Next in AI: In a surprising reversal, OpenAI is shutting down its platform for the viral text-to-video model Sora, signaling a strategic pivot to concentrate on its core business. The move comes just months after a high-profile launch that captivated the creative and tech industries.

Explained:

- OpenAI's decision is part of a larger push to refocus resources amid an increasingly crowded AI video market and a strategic shift toward developing more capable agentic models.

- The pivot involves scrapping a major deal with Disney, which included a planned $1 billion equity investment that will now not proceed following the platform's discontinuation.

- The move shows OpenAI is prioritizing its long-term roadmap over maintaining a viral product, even after the company recently achieved a $730 billion valuation.

Why It Matters: This abrupt shutdown shows that even top AI labs are making tough choices to focus their efforts in a rapidly evolving landscape. It serves as a reminder that hype alone doesn't guarantee a product's survival in the competitive AI race.

Next in AI: Chip design leader Arm is jumping directly into the AI hardware race, unveiling its first-ever production data center processor, the AGI CPU. This new chip is engineered specifically to power the next generation of AI agent applications.

Explained:

- The move targets a key shift in the AI landscape, where the focus is expanding from just training models to running them for real-world tasks, making efficient CPUs more critical than ever.

- Arm claims its new chip delivers over twice the performance per rack compared to traditional x86 platforms and is already being deployed by tech giants like Meta and OpenAI.

- Built on TSMC's 3nm process, the AGI CPU packs up to 136 cores and is already being integrated into new server platforms designed for high-density AI workloads.

Why It Matters: This launch marks a fundamental change in Arm's strategy, positioning it as a direct competitor in the high-stakes AI chip market. It also underscores a broader industry pivot toward creating specialized hardware for a future where AI agents, not just models, drive computation.

Next in AI: Google Research has introduced TurboQuant, a new set of algorithms that drastically compresses large language models with zero accuracy loss. This breakthrough promises to make powerful AI much more efficient and accessible.

Explained:

- The algorithm compresses an LLM's key-value (KV) cache to just 3 bits, which reduces memory requirements by over 6x while speeding up inference.

- It achieves this with two core components, PolarQuant and QJL, that work together to shrink model data without needing any fine-tuning.

- TurboQuant has demonstrated top performance on major benchmarks like LongBench and excels in use cases like high-speed vector search.

Why It Matters: Making AI models smaller and more efficient directly lowers the cost of running them, reducing a major barrier to wider adoption. This level of optimization could soon allow powerful, long-context models to run on consumer-grade hardware, unlocking new on-device applications.

GitHub updated its data usage policy, and will now use interaction data from Copilot Free and Pro users to train its AI models unless they opt out.

Meta hired former VR head Hugo Barra and his team from the AI agent startup Dreamer as part of a licensing deal to bolster its agentic AI efforts.

Emerald AI demonstrated in a collaboration with NVIDIA that "power-flexible" AI factories can autonomously adjust their energy usage to help stabilize public power grids during peak demand.

U.S. Senator Bernie Sanders introduced a bill with Representative Alexandria Ocasio-Cortez to place a federal moratorium on new AI data center construction, citing concerns over energy consumption and environmental impact.