ANTIGRAVITY LABJP
Articles/Tips & Best Practices
Tips & Best Practices/2026-08-15Beginner

When Antigravity Holds a Dozen Projects: How I Stopped Hunting for Yesterday's Conversation

As Antigravity projects pile up, the conversation sidebar turns into a morning scrolling ritual. Now that 2.8.0 remembers collapsed state, here is how I reworked project granularity, conversation naming, and launch flags.

Antigravity353Workspace2Productivity5Setup2

The conversation I want first thing in the morning is almost always yesterday's. Yet opening the sidebar meant scrolling past a column of near-identical names, sometimes two or three passes before landing on the right one.

I work as an indie developer with app projects and site projects running side by side, so Antigravity holds a dozen-plus projects at any given time. Growth is a good sign, but every morning began with the same twenty or thirty seconds of friction before any actual work started.

Version 2.8.0, released on August 12, made the sidebar remember its collapsed and expanded state across sessions (2.8.1 followed on August 13). Arranging things finally sticks instead of resetting on the next launch, so I took that as the moment to rethink how I hold projects in the first place.

What I was hunting for wasn't the conversation

Breaking the problem down, that scrolling time wasn't spent looking for a conversation. It was spent remembering which project it lived under.

SymptomWhat was actually happeningWhat fixed it
Can't find the conversationProjects don't match units of workRedefine granularity
Names all look alikeAuto-generated titles from the first messageFix a three-part naming rule
Conversations keep multiplyingStarting fresh threads in the wrong projectName the target at launch

Three separate-looking symptoms, one shared root: the unit of work was never defined. So I worked through them in order.

Set granularity by "what you return to," not by repository

My original split was one project per repository. Tidy on paper, but out of step with how I actually come back to work.

On the wallpaper app side, classifying image assets and preparing an App Store release live in the same repository, yet they demand completely different mental modes. On the site side, the opposite happens: for weeks at a time I'm making the same category of decision across several repositories at once.

So I reduced the rule to two questions:

  1. If I don't open it in a given week, it doesn't stay as a permanent project
  2. If I'd want to read the same conversation history, it belongs in one project even across repositories

Running everything through that filter cut my permanent projects roughly in half. Nothing needs deleting — anything retired can be reopened the moment it matters again.

The second question surprised me more than the first. I had assumed repository boundaries were the natural seams, but the history I actually reread is organized around decisions, not directories. When I split asset classification and release preparation into two projects despite the shared repository, the conversations inside each one finally read as a continuous thread rather than an interleaved log of two unrelated concerns.

With state preserved, keep exactly one project expanded

This is where 2.8.0 pays off most. Previously I could tidy the ordering, then reopen the app to find everything expanded again. The tidying never paid for itself, and at some point I quietly stopped bothering.

Now the arrangement after Collapse All survives. That let me commit to a single rule:

Only one project stays expanded at a time.

Whatever I'm focused on stays open; everything else stays folded. Switching means folding one and opening the next. That alone brought the sidebar back inside a single screen.

Worth noting: 2.6.0 (August 7) improved both the speed of opening long-history conversations and the loading of the conversation list. Folding and reopening became cheap enough that this rule stopped feeling like extra work.

Keep conversation names to three parts

Auto-generated names summarize your opening message. Reasonable in principle, but similar requests produce similar titles. Three threads that began with "help me trace this error" are indistinguishable a day later.

So I rename each conversation right after starting it, following one rule: date, subject, verb.

0815-wallpaper-categorizer-triage
0815-labs-sitemap-fix
0814-ios27beta-verify

Leading with the date means alphabetical order roughly matches chronological order. The subject can be whatever you'd call it out loud. The verb goes last so that "triage" and "fix" on the same subject sit next to each other as distinct threads.

Every additional rule is a rule you eventually skip. I keep this one capped at three parts on purpose.

One caveat worth mentioning: renaming has to happen immediately, while you still remember what you were about to do. Renaming at the end of a session sounds tidier, but by then the thread has usually wandered somewhere else, and you end up naming the last thing you discussed rather than the reason you opened it. I'd rather have a slightly inaccurate name written in the first minute than an accurate one that never gets written.

Name the target at launch instead of at the keyboard

Most of my accidental new threads came from typing into whichever project happened to be open. That's fixable at launch rather than through discipline.

Version 2.7.1 (August 11) added --project and --new-project flags, letting you state which project to open or create one outright.

# Alias the places you return to most
alias agy-wall='antigravity --project ~/dev/wallpaper-app'
alias agy-labs='antigravity --project ~/dev/labs-site'
 
# Isolate throwaway experiments from the start
alias agy-lab='antigravity --new-project ~/tmp/scratch-$(date +%m%d)'

Short experiments started inside a real project are what clutter the list later. Putting disposable work somewhere disposable from the beginning means there's nothing to clean up.

Also in 2.8.0: if you have Antigravity IDE installed, a button now opens the current project directly in the IDE. If you move between editor and agent frequently, that removes one round of reopening the same project twice.

What I left unsolved

To be straightforward about the limits: this didn't make everything comfortable.

I haven't found a way to move a conversation between projects after the fact in my setup. Threads I filed wrong simply stay where they are. That constraint is precisely why the naming rule comes first — with consistent names, a slightly misplaced thread is still findable by search.

Cross-project conversation search also doesn't feel strong yet. Rather than force a tooling fix, I now write "what did I decide last week" notes outside the conversation entirely, as short memos. It draws a clearer line between what the agent holds and what I keep at hand.

Neither limitation is severe enough to undo the change. Both are worth knowing before you invest an afternoon in reorganizing, because each one quietly shifts where the effort belongs — into naming and into launch habits, rather than into filing.

If the display settings themselves are the sticking point, it's worth reading alongside Setting Up Antigravity 2.0 with a Japanese UI. And if you want to rework how you review agent output too, The Review Routine I Rebuilt Around 2.8.0's View Limits covers how diff review changes once those limits are in place.

One thing to try tomorrow morning

Trying to reorganize everything at once rarely survives the week. Start by pressing Collapse All and choosing the single project you'll keep expanded tomorrow. Everything else can be reopened one at a time, as it becomes relevant.

Trim these rules to fit your own setup — that's how they're meant to be used.

Share

Thank You for Reading

Antigravity Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • Copy-paste ready implementation code
  • New advanced guides published daily
  • $5/mo or $10 for lifetime access
View Membership →

If you found this article helpful, a small tip ($1.50) would mean a lot to us. Your support helps keep this site ad-free and covers server and hosting costs.

Related Articles

Tips2026-07-07
When Antivirus Blocks Antigravity's Launch, Verify the Signature Before You Allow It
When the Antigravity desktop app won't launch because antivirus stepped in, here's how to tell a false positive from a real threat. Verify the code signature on macOS Gatekeeper and Windows Defender first, then allow-list with the narrowest possible scope.
Tips2026-07-05
When Your Agent's Commits Pick Up Junk Files: Fixing It With Staging Scope and a Message Convention
Agents tend to run git add -A, sweeping .bak files and caches into your history, and leave a one-word message. Here is how a staging allowlist, a preflight, and a fill-in message template stop it.
Tips2026-07-01
Don't Let Your Automation Lean on AI Ultra's 5x Ceiling
The $100/month AI Ultra plan raises Antigravity's usage limits to 5x AI Pro. But if you architect automation around that ceiling, it collapses the moment you drop back a tier. Here is a limit-independent degradation design, with the real pain points.
📚RECOMMENDED BOOKS
Build a Large Language Model (From Scratch)
Sebastian Raschka
LLM Dev
Prompt Engineering for LLMs
Berryman & Ziegler
Prompting
AI Engineering
Chip Huyen
AI Eng
* Contains affiliate links
See all →