In this video I tested Google Antigravity and honestly… it feels like the future of coding with AI. I walk you through the full setup on a local machine (I’m on Windows): download Antigravity, create a project folder, and use the Agents tab + Agent Manager so you can run multiple agents in parallel. The big unlock is that inside the IDE you can pull in different models (I start with Gemini 3.1 Pro on high reasoning + planning mode) and let the agent generate a real Next.js app from a simple markdown file that describes what you want. The app I build is something I’d actually use for my AI startup: a competitor pricing comparison tool. I integrate Firecrawl to scrape competitor sites, then format the messy scraped data into clean, visual comparison tables. Along the way, you see the real workflow: the agent asks clarifying questions (framework, database, API key), generates a plan, builds the project, runs the dev server, and even uses the built-in browser to test and self-correct. I hit real issues (wrong Firecrawl endpoint, async job timing), pivot the strategy to pricing pages only, and fix polling—plus I switch models to Opus 4.6 when my Gemini credits run out. The takeaway: with Antigravity + the right model + a tool like Firecrawl, you can ship useful internal SaaS-style tools fast.