Anthropic has announced that it is making the 'auto mode' for its AI agent, Claude Code, the default setting. This change aims to significantly enhance efficiency for developers by allowing Claude to autonomously execute coding tasks.

With auto mode, the LLM can continuously generate code, debug, and suggest improvements without requiring explicit step-by-step prompts from the user. This represents a significant step towards AI agents managing more complex workflows and accelerating the development process.

👉 Read the full article on TechCrunch

Concerns are being raised that current methods of AI safety testing are inadvertently creating new safety risks. The article points out that these tests could be exploited by malicious actors or might overlook more subtle or emergent dangers by focusing on specific, measurable risks.

Methodologies for evaluating AI models, especially LLMs, for bias, misuse, or unintended consequences are being critiqued, suggesting that the tests themselves might be insufficient or even counterproductive in ensuring true AI safety.

👉 Read the full article on TechCrunch

Discovered Materials is leveraging AI to accelerate the discovery of new materials, specifically those that can lead to more efficient and cooler-running computer chips. This 'whack-a-mole' approach suggests an iterative, trial-and-error process that is significantly sped up and optimized by AI algorithms.

Generative AI and advanced machine learning models are being used to predict material properties, simulate performance, and guide experimental synthesis, thereby reducing the time and cost associated with traditional materials science research. The goal is to find novel compounds that can dissipate heat more effectively, which is crucial for the next generation of high-performance computing and AI hardware.

👉 Read the full article on TechCrunch