ThursdAI - The top AI news from the past week · From Weights & Biases, Join AI Evangelist Alex Volkov and a panel of experts to cover everything important that happened in the world of AI from the past week

ThursdAI Special: Google's New Anti-Gravity IDE, Gemini 3 & Nano Banana Pro Explained (ft. Kevin Hou, Ammaar Reshi & Kat Kampf)

·46 min·2 clips

Kevin Ho defines artifacts as dynamic outputs like task lists, screen recordings, and curl requests.

1. ThursdAI Special covers Google’s November 18 releases of Gemini 3, Anti-Gravity IDE, and Nano Banana Pro around the AI Engineer Code Summit in New York. 2. Alex Volkov speaks with Kevin Ho from Google DeepMind’s Anti-Gravity team, then with Amar Reshi and Kat Kampf from Google AI Studio, because each team ships a different part of Google’s developer stack. 3. The episode asks what builders can do with Gemini 3, Anti-Gravity, and AI Studio when multimodality, coding agents, and vibe coding arrive in the same week. 4. Kevin Ho says Anti-Gravity is a new agentic IDE based on VS Code and says the team shipped it during a week that also included Gemini 3 and Nano Banana Pro. 5. He says this is his first Google product launch and describes the launch week as “surreal,” with a 6 a.m. release and strong early feedback. 6. Kevin Ho describes a test where the agent planned a feature, wrote the artifact system, generated images, opened the app in a browser, and then produced a screen-recording walkthrough. 7. He says that moment made the team realize Gemini 3 plus Anti-Gravity could do something that had not been technically possible before. 8. Kevin Ho defines artifacts as dynamically generated outputs that can be implementation plans, screen recordings, curl requests, or task lists stored for later use. 9. He says artifacts are dynamic rather than files on the user’s machine, and that the model can decide which artifact form fits the task. 10. Kevin Ho explains that artifact comments work like Google Docs, because users leave comments while the agent is still running and the agent folds the comments into execution. 11. He says this shifts the workflow from watching raw diffs to supervising agents through screenshots, video, and notification-based attention. 12. Kevin Ho says Chrome integration matters because computer use is a frontier capability and the Anti-Gravity team worked closely with the Chrome team for months. 13. He says the browser sub-agent can click, drag, scroll, run JavaScript, and annotate screenshots when the browser surface needs more than simple control. 14. Kevin Ho says Anti-Gravity stores artifacts such as architecture diagrams, API documentation, and verified curl outputs so later tasks do not need to rediscover them. 15. He says the biggest missing piece is reliable research over Google’s large internal corpus, especially for the company’s biggest codebase and documentation set. 16. Kevin Ho also says mobile development and desktop applications remain short-term gaps for computer use. 17. Amar Reshi says AI Studio is Google’s home for builders and developers who want a Gemini API key, chat with new models, and build apps quickly. 18. Kat Kampf says a new user now gets an API key automatically, while existing users can deploy to Cloud Run and keep control of the URL and billing. 19. The conversation is technical but conversational, with live-demo energy and many product details surfaced through back-and-forth examples. 20. People building with Gemini 3, AI Studio, or agentic IDEs would likely get the most value, while listeners wanting nontechnical product talk may skip it.

As heard by us

A brisk tour of Google's Anti-Gravity IDE and the AI release week around it.

Google's release week turns into a brisk look at its newest AI coding tools, led by Kevin Ho's walk through the Anti-Gravity IDE and its Agent Manager, Artifacts, and still-visible limits in model behavior.

Read the full review in PlayNext →

Why you'd press play

Want Gemini 3, the Anti-Gravity IDE, and Nano Banana Pro decoded from the conference doorway before the bananas multiply?

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from ThursdAI - The top AI news from the past week