Introducing Glance, a fast browser agent trained with RL to test code changes. We're releasing this with BrowserBot, which puts a video of it testing your app right in your GitHub PRs Frontier models are trained to be helpful. Our model is trained to be bad at using software.
Nobody reads PRs anymore. AI writes 5000 lines. Reviewer scrolls for 10 seconds We embed a video of an agent actually using your app's preview URLs right in the PR. Now you can watch instead of pretend to read.
Why train this? We turn code testing into a search problem – searching for components that changed via browser tools. The hard part is grounding. A diff in Button.tsx doesn’t tell you where that button lives or how to reach it. We walk React’s Fiber tree to map changed files →
The reward: • Δ coverage when changed component enters viewport • 2x reward for interacting with it • Reward for novel state discovery
Humans were only in the loop cause code review sucked. Now i can go play league of legends
Don't miss what's happening
People on X are the first to know.
