Tech • AI • Robotics • Game
Google Anti-Gravity meets Base44 Astra as AI coding goes full-stack
Briefing
Google launches Anti-Gravity IDE
Google-backed Anti-Gravity emerged as a new coding environment built on VS Code, aimed at turning plain-English requests into working application changes. The product behaves like an in-editor agent rather than a simple autocomplete layer, with a dedicated prompt pane alongside the standard editor, file tree and terminal. It is positioned to plan tasks, modify multiple files and execute commands inside a bounded workspace. That makes it a direct play at the growing market for agentic developer tools inside familiar IDEs.
Project permissions gate the agent
The biggest operational constraint is project setup: the agent does not act until a folder is explicitly linked and granted access. If that step is skipped, prompts appear to fail even though the real issue is missing permissions. The model uses that project boundary to determine which files it may inspect or change. In practice, permissions and folder linking are the critical onboarding hurdle for new users.
Anti-Gravity emphasizes plan-then-approve flow
A notable design choice is a plan-first workflow in which the agent outlines intended actions before code is altered. Users can then approve the proposed implementation, adding a review step before file edits and terminal operations begin. That approval gate is increasingly central to enterprise-friendly AI coding products seeking auditability and control. It also reframes the tool as a collaborator with explicit checkpoints rather than an always-on code generator.
Node.js to-do app proves workflow
A simple Node.js to-do application was used to demonstrate the end-to-end loop from prompt to generated files. The target feature set included adding tasks, marking them complete and deleting them through a clean interface. Once the folder was linked, the agent generated a plan, waited for approval and then created the required code. The example was basic, but it showed the product handling multi-step app scaffolding inside the editor.
Google Cloud deployment is baked in
Anti-Gravity is also framed as a path toward deployment, with preparation for Google Cloud highlighted as part of the workflow. That suggests Google is tying code generation to its infrastructure stack rather than treating the IDE as an isolated assistant. For developers, the appeal is a more continuous path from prototype to hosted service. For Google, it extends the contest over AI coding into cloud capture and platform loyalty.
Base44 and Astra build Card Vault
On the no-code side, Base44 paired with Astra was used to create Card Vault, a mobile-first collector app for trading cards. The concept targeted Pokémon, MTG, baseball and basketball cards, giving the build a narrow but commercially recognizable niche. The app combined database records, user accounts and a polished mobile interface in a rapid assembly flow. The result illustrates how AI builders are pushing beyond toy demos into category-specific consumer products.
Card Vault adds search and AI scanning
The app’s core data model covered fields such as name, card type, collection, category, condition and notes, establishing a structured vault for each user. On top of that, the build layered live search, filter pills, a condition dropdown and AI image scanning for card recognition. Wish lists and grading-style indicators expanded the product from simple storage toward discovery and collection management. That feature mix is significant because it compresses database design, UI logic and AI inference into a no-code workflow.
Hosted publishing gets the spotlight
A key takeaway from Base44 and Astra is not just app generation but immediate live hosting and account-enabled publishing. The emphasis was on shipping a usable, mobile-oriented product quickly rather than only prototyping screens. This highlights a broader split in AI coding: editor agents like Anti-Gravity target developers inside codebases, while no-code builders target end-to-end product delivery. Together, the two launches show the market converging on full-stack automation from prompt to deployment.
Videos covered
How Complete Beginners Master 85% of Antigravity in 19 minutes
- •Project setup is the main beginner hurdle
- •The tool works as an in-editor agent, not just autocomplete
- •A simple Node.js to-do app demonstrates the workflow
GPT 6 Astra + Base44 = $10,000 Apps in Minutes
- •A niche app as a stress test
- •Core collection system came first
- •Filtering and search were layered onto the vault