In a candid interview with The New York Times, Google CEO Sundar Pichai acknowledged that the company is currently trailing the cutting edge in AI-powered programming, tool calling, and long-horizon tasks. His explanation? Google lacks a direct entry point into developers' daily workflows — something like Cursor or Claude Code — and as a result, it's losing the most critical dynamic workflow data.
https://www.youtube.com/watch?v=RgV57kDzcng
But Pichai's admission about a "missing public entry point" doesn't quite line up with reality. Google actually shipped the first version of its AI coding environment, Antigravity IDE, back in November. The rollout was a mess: quota bugs locked users out, servers repeatedly buckled under load, and restrictions on third-party open-source API integrations drove early adopters straight into competitors' arms.
During the interview, Pichai highlighted impressive internal metrics from Antigravity running inside Google's own walls — token consumption doubling every week, and a stressed test that built an entire operating system from scratch in 12 hours. The message was clear: the underlying tech is solid; the public-facing failure is a problem of surface area and feedback loops.
Last week, Google rushed out Antigravity 2.0 in open beta, hoping to flex its technical muscles and win back developers with better terms. The big interview push — boasting internal wins while admitting the absence of a user entry point — is really about setting the stage for 2.0's comeback and finally closing the loop on real-world developer feedback and algorithmic iteration.