What is Antigravity?
The Harvard Chan AI Glossary
In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.
Antigravity is Google’s agentic development platform: a working environment for building software with AI agents. An agent is software that can select actions toward a goal and adjust its next step after observing the result. A development environment brings together the project files and functions needed to create and examine a program.
The important distinction is between the environment and the model. A model is the trained component interpreting instructions and proposing responses. Gemini names a model family; Antigravity names a product around model capabilities and development tools. Choosing one does not make the two names interchangeable.
An illustrative task is adding a search function to a website. An agent may inspect existing files, propose changes, run the program and examine the result. Code is the written set of instructions a computer executes. The environment gives the agent ways to read and change that code, subject to its configuration and permissions.
That is different from receiving a code suggestion in a chat and copying it manually. The working environment can support a sequence of actions, but the user still needs to specify the intended behavior and decide what counts as success. A running page does not establish that every search case works correctly.
Review matters because generated changes can affect more than the visible feature. An altered file may change how information is stored or who can access it. A test is a check of expected behavior under stated conditions. Looking at the proposed changes and examining test results provide different kinds of evidence, both of which may be needed.
The available interfaces and model combinations change. Check the documentation for the version you are using, particularly its controls for file access, command execution and review.
The useful question is what the configured environment lets the agent inspect, change, run and verify. A familiar model name does not answer those questions. Antigravity belongs in the glossary as a development platform, not as another name for the intelligence of its underlying model.
Source: Google — Antigravity Overview.
If you would like to find out more about AI and improve your leadership skills, consider the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.
AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.
For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.
Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy