> [ GENEVA, SWITZERLAND — March 28, 2026 ] — CERN is using extremely small, custom large language models physically burned into silicon chips to perform real-time filtering of the enormous data generated by the Large Hadron Collider (LHC).