11 related articles

In-depth review of Google's Antigravity 2.0 desktop Agent app, testing Gemini 3.5 Flash code generation, scheduled task automation, and dynamic Sub-agent parallel collaboration features.

Antigravity 2.0's redesigned IDE faces massive backlash: frequent bugs, UX regression, limited model support, abnormal Gemini Token consumption, and reports that its own developers don't use the product.
产品体验Antigravity 2.0 moved its IDE feature entry to a prominent upper-right position via UI update, resolving user confusion about the feature being removed, while resetting weekly quotas as compensation.
产品体验Google AI Studio integrates the Antigravity coding agent with deep Firebase support for databases, auth, secrets management, real-time sync, and Next.js — pushing Vibe Coding toward production-grade full-stack development.

A Google DeepMind engineer reveals that over 50,000 AI agent skills come with almost no evals. Learn how to write skill evaluations, from description tuning to test design.

A Google DeepMind engineer reveals that over 50,000 AI agent skills come with almost no evals. This guide covers skill descriptions, test design, eval harnesses, and retirement strategies.

OpenSpiel 2.0 by Google DeepMind adds LLM fine-tuning examples, MCP tool server, JSON trajectories, AlphaZero on JAX, 19 new games, and Windows support.

July 12 GitHub trending: Agent Skills/MCP ecosystem explodes with superpowers hitting ~900 stars, pgrust rewrites Postgres in Rust passing 100% tests, plus solid engineering foundations.

Google launches unified AI platform Antigravity, migrating Gemini CLI users to the new Antigravity CLI rebuilt in Go with multi-agent orchestration and async workflows.

Deep dive into Google Firebase's integration with AI Studio, covering four core capabilities—database auto-configuration, authentication, security rules drafting, and zero-config deployment—for production-grade AI agent apps.
产品体验Deep dive into Google's Antigravity IDE: analyzing this free AI coding tool built by the Windsurf team, its agent-first development mode, real-world performance, and full comparison with Cursor.