Lesson 1.2 · Module 1, Fundamentals
The two views and the artifact trail
Editor view for steering, Manager view for parallel work, and why the artifacts an agent leaves behind are the thing you should be reading.
Antigravity gives you two ways to work, and choosing wrong is the most common reason people find it slow.
Editor view: when you want to argue
A conventional editor with your files on the left, the document in the middle, and an agent panel on the right. You write, it responds, you push back, it revises. Use it when the work is a conversation: drafting, cutting, re-framing, deciding.
The interaction that matters here is the @ mention. Typing
@context/company.md attaches that file rather than describing it. Attach
three files and the agent has your brief, your personas and your transcript. Attach the
whole folder and it has nothing in particular, because everything is equally weighted.
Manager view: when you want to leave
Google describes Manager view as a control centre for orchestrating multiple agents working in parallel across workspaces, with tasks running asynchronously. In practice you send off several long jobs and go to a meeting.
Good candidates: summarise forty support tickets; produce a competitor comparison from six saved pages; turn a transcript into a themed digest. All of them are long, none of them need you.
Bad candidates: anything where you would want to interrupt after the second paragraph. Sending a PRD draft to Manager view produces a finished document built on an assumption you would have caught immediately.
| The work | View | Why |
|---|---|---|
| Drafting a PRD section | Editor | You will change direction three times |
| Summarising 40 tickets | Manager | Long, mechanical, no decisions |
| Rewriting a paragraph you dislike | Editor | Faster to say than to specify |
| Comparing six competitor pages | Manager | Parallel and slow |
| Deciding what to cut from scope | Neither | That one is yours |
Read the artifacts, not the transcript
Instead of showing every tool call, Antigravity surfaces artifacts: a task list, an implementation plan, screenshots, a browser recording. This is the review surface, and most people ignore it and then wonder why they were surprised by the output.
Three habits make it worth the space it takes:
- Read the task list before the result. If the tasks are wrong, the output cannot be right, and you have saved yourself reading three pages to discover it.
- Check which files it actually opened. If it produced a competitive analysis without opening your competitor notes, it wrote that from training data. That is not analysis, that is a plausible guess.
- Keep the recording when it browsed. If the agent went and read a pricing page, the recording is your evidence for the number it quoted. Screenshots age better than claims.
Exercise
- In Editor view, ask: “Read
@context/company.mdand list the three assumptions in it that you would most want evidence for.” - Now open Manager view and send the same request as a background task.
- Compare. The Editor answer arrives faster and you will have pushed back twice by the time the Manager task finishes. Note which one you would actually use, and for what.
The point of the exercise is not that one is better. It is that you now have a felt sense of the cost of each, which is what stops you defaulting to whichever is on screen.
Independent and unofficial. This site is published by readers of Antigravity, not by Google. It is not affiliated with, endorsed by or connected to Google LLC. Nothing is hosted here — no software is mirrored, repackaged or distributed, and no installer, licence key, sign-in or payment is ever requested. Antigravity itself is downloaded only from Google, at antigravity.google. Never enter a Google password, a payment card or an API key on this site — nothing here needs one.