Why I reach for Antigravity (Cursor series Part 2)

Part 2 of a series.

Part 1 said the quiet part: Cursor isn’t only where I write code. It’s where job search, drafts, fit notes, and a named roster live, with me on the dangerous verbs.

So why open Google Antigravity at all?

Because sometimes the job is “spin up agents, let them chew a codebase, review artifacts,” and I don’t need the whole studio kitchen for that. Antigravity is built as an agent-first surface: manager / mission-control vibes, parallel local agents, IDE when you want the code under your thumbs. Cursor remains where the studio’s operating system lives. Different instruments.

When Antigravity wins

Parallel coding tasks with a clear blast radius. Refactor this package while tests get green elsewhere. Scout a UI path in a browser-in-the-loop agent while another agent stays in the repo. When the work is software-shaped and I want concurrency without inventing my own orchestration ritual in chat, Antigravity’s manager surface is the point.

Gemini-lane coding without relocating my life. Some days the model fit is Gemini-heavy and the task is “build / verify / leave artifacts I can inspect.” I dual-wield the way Google itself suggests: agent command center plus an IDE I already trust. For me that second surface is often still Cursor, or Antigravity’s own IDE when I’m already deep in that session.

Throwaway spikes that shouldn’t pollute the studio ledger. Researchy “can this even compile” work. A disposable workspace. I don’t want every exploratory agent run sitting next to job-search upserts and Notes drafts in the same mental window.

When I don’t reach for it

Anything that isn’t primarily code. Job-hunt CLI verbs, resume diffs against originals, research briefs Page drafts (that’s editorial work, not code), stays on the Cursor roster. Antigravity is not my CRM for applications and it’s not my Notes kitchen.

Work that needs my house rules as first-class routing. Named agents with phase gates (discover before dig, hygiene before features, Tesla on the pipeline and integration edges). I already designed that job chart. Porting the whole crew into a second product every morning is not productive. It’s theater.

Moments when I need the lighter hammer. Tab complete, small edit, VS Code + Copilot. Part 3 will cover that middle. Antigravity is not “open this to rename a variable.”

Anything where unsupervised autonomy would feel clever. Same rule everywhere: humans on push, confirm before mutate, secrets get a backup path. A shinier agent manager does not earn unsupervised egress. Security vs theater still applies.

Dual-wield, not divorce

People hear “series about Cursor” and assume I’m picking a religion. I’m picking a default, and calling in specialists when the job actually needs one.

Cursor stays the control surface. Antigravity gets the invite when parallel, agent-first coding is the whole point of the hour. Copilot steps in when I don’t need a crew at all, just a lighter hammer for a small edit.

Vendors will keep shipping mission-control metaphors. My job is to notice which metaphor matches the work, not which logo is loudest on X this week.

What’s next

VS Code + Copilot as the quick/painless middle. Then a deeper cut on agents in the job hunt if the series still has oxygen.

For now: Antigravity is the cousin I call for movie night when it’s their favorite genre. Great company for the right hour. Not someone who gets a key to the studio.

~ Trish