Anthropic AI researcher Andrej Karpathy has suggested using long voice input rather than crafting precise prompts when working with generative AI. He said he often turns on voice mode and speaks freely for about 10 minutes about a project, letting the model organise the stream of thoughts. He said this can better align context and reduce later revisions, though some X users were sceptical.

Artificial intelligence is becoming core infrastructure in biopharmaceutical development, rapidly changing how drug candidates are found and validated. AstraZeneca says computing now reinforces the entire process from design and production to verification and analysis, shortening development cycles and boosting productivity. The company uses a build-measure-learn loop, multimodal datasets and precision screening, and is building an AI- and robot-enabled lab in Cambridge, Massachusetts. Executives say de novo, fully AI-generated biologics are an industry goal but require richer data, benchmarks and talent, with human oversight remaining central.