Google Goes All-In on AI Coding: Antigravity Platform, Gemini 3 Flash & Pro in CLI — Everything New

TamilTech AI summary

Google just rolled out a major AI coding upgrade with Antigravity, a free public-preview agentic platform that pairs a familiar AI IDE with a Manager Surface where you can spawn multiple agents to write code, run terminals, test UIs in the browser, and send back Artifacts like plans and screenshots you can comment on. Gemini 3 Flash landed in Gemini CLI with a strong 78% SWE-bench score—beating even Gemini 3 Pro on that benchmark while costing roughly a quarter as much—and smart auto-routing picks Flash for routine work or Pro for deeper reasoning. Gemini 3 Pro was already in the CLI for heavy tasks like building complex apps from one prompt, chewing through huge codebases, and running full agentic loops with shell, files, and docs. This matters because Google is shifting from “AI that helps you code” toward “AI that builds software for you,” putting it in direct competition with Copilot Workspace, Cursor, and OpenAI’s coding agents. Developers should know Antigravity is free to try on macOS, Windows, and Linux, Gemini CLI upgrades are straightforward for paid/API/AI Pro/Ultra users, and while these tools are still in preview with rough edges, the direction points to AI agents becoming a normal part of professional workflows.

• What is Google Antigravity?
• What is the SWE-bench score for Gemini 3 Flash?
• How does Gemini CLI's auto-routing work?
• What are some capabilities of Gemini 3 Pro in Gemini CLI?

AI-assisted summary, checked by the TamilTech editorial team.