Google's new platform, named Antigravity, enables AI agents to act more like full-fledged software developers rather than just coding assistants. On several supported operating systems, Windows, macOS and Linux, these agents can access tools including code editors, terminals and web browsers to build and validate applications.

In one demonstration, an agent built a flight-tracker app completely using Antigravity while logging its steps through what Google calls “Artifacts”, plans, screenshots and browser recordings that document its work.

Coding Model Behind The Platform

Antigravity is powered by the Gemini 3 model at its core. Google describes Gemini 3 as its most advanced model yet, excelling at reasoning, multimodal input and extended tool use including direct terminal commands.

By shifting the paradigm from “assistant helps developer” to “agent executes tasks under developer supervision”, Google signals a next phase of how code may be generated and validated in enterprise and consumer software workflows.

Software development has long leaned on human coders navigating editors, compilers and debugging tools. With Antigravity, the human coder shifts toward oversight and review rather than step-by-step implementation. That raises both opportunities and questions.

Opportunities include faster development cycles, more prototype generation and leveraging AI to reduce repetitive coding tasks. Questions centre on how trust, audit and control will work when agents act more independently, for example, how the chain of responsibility or version control adapts when code is authored partly by AI.

Earlier tools in software generation, such as code-completion assistants, focused on local suggestions inside the editor. AI agents that control full workflows remained experimental. Google’s announcement ties into a larger trend where major tech companies integrate agentic workflows into development platforms.

Competitors and prior launches have hinted at that shift. Yet Antigravity represents one of the first broad-scale offerings where agents can operate across multiple tools, not just inside an editor. That may accelerate the transition of AI from support role into co-creator role in development environments.

What To Watch Going Forward

How developers and teams adopt Antigravity will be a key indicator, whether it serves best in prototyping, production or entirely new workflows remains to be seen.

The robustness of the “Artifacts” logging and agent insight features will shape trust and governance models in enterprise software.

Also relevant will be how software tools around auditing, security and code review evolve to account for agent-generated contributions. If this model works, we may see new norms in how code is authored, reviewed and maintained.