gemini-3.8-flash
Google's Agents CLI Adds an Eval Loop to Catch Agent Regressions
Google has published a walkthrough of the eval namespace in Agents CLI, a set of commands for testing, grading and comparing AI agents so that teams can measure whether…
AI Taste Study: Spending 9.5 Times More Bought Only Four Points of Preference
Paying more for inference did not reliably produce web designs that people preferred, according to a new analysis from HumanSignal, the company behind Label Studio. In…
Ten AI labs, one orange: models converge on the same palette for a skate brand
Given a brief for a New York street-skate brand and no visual direction, nine of ten AI models shipped a safety-orange accent in their built sites, according to…