Tech • AI • Robotics • Game
Google Anti-Gravity lands as Typesafe AI launches Jev
Briefing
Google unveils Anti-Gravity IDE
Google-backed Anti-Gravity emerged as a new coding environment built on VS Code and aimed at prompt-driven app development. The product acts as an in-editor agent rather than a simple autocomplete tool, using plain-English instructions to plan tasks, edit files, and run terminal commands. Its positioning puts it squarely in the fast-growing market for agentic developer tools. The workflow is oriented toward turning natural-language requests into directly executable project changes.
Anti-Gravity makes project permissions central
A defining design choice in Anti-Gravity is its requirement that users explicitly create a project and link it to a local folder before the agent can operate. That permission boundary determines which files the system may access, and missing the setup can leave prompts apparently unresponsive. The approach emphasizes controlled scope over instant access. It also mirrors a broader industry shift toward tighter governance in coding agents.
VS Code agent edits whole apps
Inside VS Code, Anti-Gravity runs through a side panel while preserving the standard editor, file explorer, and terminal layout. The agent can outline a plan, seek approval, modify multiple files, execute commands, and then summarize progress. That separates it from tools centered on inline code suggestions. The result is a more autonomous development model embedded inside a familiar interface.
Node.js demo shows deployment path
A basic Node.js to-do application illustrated how Anti-Gravity moves from prompt to working code. After project linking, the agent generated a plan, waited for user approval, and then created and updated the necessary files. The flow highlighted iterative app building rather than one-off code completion. Preparation for deployment to Google Cloud is part of the product’s intended arc.
Typesafe AI launches Jev
Typesafe AI introduced Jev, an early-access decision engine built for structured outputs rather than conversational text. The service launched on September 15, 2026 and is being distributed via a waitlist at typesafe.ai. Its pitch is speed and operational utility for workflows that need a single machine-readable answer. That places it closer to decision infrastructure than to general-purpose chatbots.
Jev targets millisecond workflow decisions
Jev is designed to return results such as booleans, scores, and fixed choices in milliseconds. Instead of generating prose that downstream systems must parse, it emits typed outputs a workflow can consume immediately. That makes it relevant for high-volume screening, routing, and moderation tasks. The product reflects demand for narrower, faster AI components inside production systems.
Jev’s three primitives stand out
The core of Jev is a trio of primitives: null, score, and choice. Null estimates how true a statement is on a 0 to 1 probability scale, score grades content against a rubric, and choice selects one option from a predefined list with confidence. The structure is deliberately constrained to support deterministic integration patterns. It is a notable contrast with open-ended response models.
Early access signals enterprise intent
Access to Jev is currently limited, with users entering through console.typesafe.ai after waitlist approval. The console includes Playground, Usage, API keys, and Documentation, indicating a developer-first rollout. Reported approval delays of roughly three hours suggest demand management rather than open public availability. Together with Anti-Gravity, the launch points to a market split between coding agents and specialized decision engines.
Videos covered
Jev Full Tutorial for Beginners: How to Use & Set up Jev (Typesafe)
- •Purpose-built for decisions
- •Early-access rollout
- •Three core primitives
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