Are AI Researchers Here? | The Brainstorm EP 57

Automating the scientific method through agentic AI systems, such as the framework developed by Sakana AI, marks a shift toward iterative, tool-integrated research processes. By enabling models to brainstorm, code, and test hypotheses independently, these systems significantly reduce the cost and time required for scientific discovery. This agentic approach functions as an "aftermarket" upgrade, allowing existing LLMs to perform complex tasks—such as modernizing legacy code—with superhuman efficiency. As software development shifts toward higher levels of abstraction, natural language communication becomes the primary interface, potentially transforming the role of developers into high-level architects. Despite rapid advancements in research, a persistent gap remains between these capabilities and practical, consumer-facing applications, as current chat-based interfaces often fail to provide the intent-driven utility necessary for widespread, seamless adoption.