Which AI writes the better take? You decide — blind.

OpenAI and Molecule.one demonstrated a near-autonomous AI chemist powered by GPT-5.4 that improved a challenging reaction used in drug development. The work points to how AI systems could help accelerate medicinal chemistry research by optimizing key steps in molecule synthesis.

Researchers proposed a self-evolving LLM-based agent that improves legal case retrieval by automatically creating, testing and pruning query-rewriting rules for BM25 without parameter training. Evaluated on the Chinese LeCaRD-v2 benchmark, the framework outperformed non-evolutionary baselines such as human-designed rules and greedy rule selection, with gains tied to the LLM’s ability to use prior experimental feedback and eliminate weak rules.

Hugging Face has introduced Agentic Resource Discovery (ARD), an open specification allowing AI agents to dynamically search for tools, skills, and other agents at runtime rather than relying on pre-installed capabilities. Developed with industry contributors like Microsoft and Google, ARD enables federated registries to catalog and index resources, improving scalability and reducing manual integration. Hugging Face’s reference implementation, the Discover Tool, already provides search access to thousands of skills and applications across its platform and other ARD-compatible services.

Google Docs users can disable Gemini prompts and other AI writing features by turning off Google Workspace “smart features” through Gmail settings. The guide frames the change as a way to stop intrusive AI pop-ups and writing suggestions from interrupting work in Docs.

Mistral has launched Search Toolkit, an open-source framework designed to simplify building production search pipelines for AI applications by unifying ingestion, retrieval, and evaluation tools. The toolkit aims to reduce engineering overhead, allowing teams to focus on improving search quality rather than maintaining integrations across disparate systems. It supports cloud, on-premises, and edge deployments, offering consistent processing for diverse data sources and built-in evaluation for retrieval performance.