Antigravity enables /teamwork-preview for paid plans, warns on tokens

Antigravity has just rolled out /teamwork-preview for all paid plans, bringing parallel implementation and verification agents for complex tasks. Varun Mohan says it’s a research preview that can burn through tokens—and claims it’s already built a working OS.

•
LLMWorkflowsOpenAI
Antigravity enables /teamwork-preview for paid plans, warns on tokens

TL;DR

  • /teamwork-preview enabled: Available on all paid plans; positioned as a research preview
  • Parallel implementation + verification agents: Designed for complex tasks; reported high token consumption
  • Availability context: Antigravity 2.0 “right now”; not in the IDE yet
  • Operational requirement: Local machine must be kept awake during runs
  • Model and plan references: Mentions Gemini models; highlights Google AI Ultra $200/mo plans
  • Quotas and overages UI: Lists Gemini/Claude/GPT-OSS quotas; includes AI Credit Overages toggle and refresh-based credits

Varun Mohan, in a post on X, states that Antigravity has enabled /teamwork-preview for “all paid plans,” describing it as a research preview that runs “parallel implementation and verification agents” for complex tasks. Mohan also claims the feature has been used to build “a working OS,” while warning that it “can consume lots of tokens.”

The attached screenshots appear to show /teamwork-preview documentation and Antigravity interfaces. One image includes language that the command was “orchestrated using the core agent primitives newly available in Antigravity today,” says it is a “research preview,” and notes that users need to keep the local machine awake. The same material also mentions use with Gemini models and highlights “Google AI Ultra $200/mo plans.”

Mohan later replied to a commenter that the feature is “available on 2.0 right now” and “not the IDE yet.”

Source: X post by Varun Mohan

Continue the conversation on Slack

Did this article spark your interest? Join our community of experts and enthusiasts to dive deeper, ask questions, and share your ideas.

Join our community

Related Articles

Ramp finds AI coding agents won’t self-limit token spend

Ramp finds AI coding agents won’t self-limit token spend

Ramp Labs says coding agents blow past budgets even with live meters and explicit approvals. In SWE-bench tests, agents almost always chose to keep spending, and separate “controller” models were easily swayed by bad recommendations.

2 shared tags
Armin Ronacher warns Pi-on-Pi dogfooding is getting messy

Armin Ronacher warns Pi-on-Pi dogfooding is getting messy

In a newly published post, Armin Ronacher digs into what happens when Pi is used to build Pi—and why LLM-shaped issue reports can add confident “slop.” He also breaks down the scale problem in trackers and argues for stronger shared foundations over patchwork fixes.

1 shared tag
openai cover

OpenAI adds Codex “Sites” to turn plans into shareable apps

OpenAI has just rolled out Codex “Sites,” bringing a new way to turn work and ideas into interactive websites or apps shared by URL. It’s launching first for Business and Enterprise users, as developers debate where it fits and how it handles security, auth, and scale.

1 shared tag