Blog Review: July 22

Synopsys’ Dana Neustadter and Ilya Tolchinsky warn that physical AI creates a different kind of security challenge that requires continuous runtime assurance to ensure outputs remain within safe operating bounds as conditions change. Cadence’s Hamid Shojaei argues that building a reliable agent for chip design relies on closing the loop, where the agent proposes something, a tool gives it real high-quality feedback, the agent reasons over that feedback and refines, and you repeat until you conve