What engineering leaders and their teams report about Google Antigravity, with every finding tied to its evidence.
Read the findingsPractitioner sentiment, all practitioners
−16 Mixed from 102 sources that take a side
How this figure is madeFrom 98 authors in 79 discussions. Use a task pilot to measure accepted work and review time.
Nov 2025 to Oct 2026.
123 sources from practitioners, by month and sentiment.
Public accounts lean negative. Most sources that discuss quotas raise a problem, and reliability draws concern from 24. Praise goes most often to multi-step execution, plan and walkthrough artifacts, and access to Claude models.
Positive, mixed and negative sources
The score counts them as 32 positive, 18 mixed and 48 negative votes, one per author. See How the numbers work.
See the evidenceBelow: 5 praise themes, 7 concerns, 2 close to agreement and 3 disagreements.
Counted for All practitioners, from 123 sources. Change group
Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.
Counted for All practitioners.
Positive, 17
Positive
taylorfinley, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
Given an error message, agy with Gemini 3.8 Flash attached GDB to their GPU driver, reverse-engineered the ioctl interface and wrote a working LD_PRELOAD shim.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
martythemaniak, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
Over SSH, agy set up Docker, CUDA, ROS2 and Foxglove for their lidar on a Jetson and it all worked, using 6 percent of their weekly limit.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
evilduck, Hacker News commenter
Comment on: Benchmark: CadQuery vs. OpenSCAD for agentic CAD workHacker News comment
Had Gemini 3.8 Flash in Antigravity swap a fan shroud design into an existing STL of a computer case front plate. It made 2D renders from each axis, which the author takes to be for its own visual check, then wrote a Python program that manipulated the STL geometries, and it went perfectly.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
allthetime, Hacker News commenter
Comment on: Gemini 3.8 Flash and 3.8 Flash CyberHacker News comment
allthetime now uses Antigravity as their main tool at work with all their company's services in one project, planning with the agents before letting them run across service boundaries.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
dpweb, Hacker News commenter, student plan
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
On a student Google AI Pro plan dpweb likes the weekly allowance and Flash 3.7, bought extra credits, and uses agy for devops such as SSH servers on their home network.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Zachary Flower, developer
Vibe Mapping: Google's Antigravity is pretty damn amazingPersonal blog post, flower.codes
Pointed at the source of an old MUD, Antigravity worked out the proprietary area-file format and built a map explorer, about 80 percent on its own in an hour.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Positive
jchw, Hacker News commenter
Comment on: How Google got its groove back and edged ahead of OpenAIHacker News comment
Running Antigravity for hours on hard debugging problems, jchw saw it break out of loops and make real progress, using gdb in batch mode on crashes.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability
Positive
scotty79, Hacker News commenter
Comment on: AI agents are starting to eat SaaSHacker News comment
scotty79 built a side-by-side HTML diff tool in Antigravity in one evening, correcting the agent repeatedly when it ignored uv and shipped template mistakes.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
Philipp Hartenfeller, Product Lead at United Codes
This blog is now on Astro thanks to Antigravity!Personal blog post, hartenfeller.dev
Antigravity with Gemini 3 Pro migrated their blog's React components from Gatsby to Astro over several bug-fixing iterations without being prompted to continue.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability
Positive
joaoviniciusgs, Hacker News commenter
Comment on: I made an open-source Boom 3D clone. Wanna check it out?Hacker News comment
joaoviniciusgs built an open-source Rust and Tauri audio app with Antigravity and Gemini 3.8 Flash, using the BMAD framework to plan, and was surprised how far structure took the model.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
Joske Vermeulen, runs an autonomous agent startup race
My AI Agent Hit Google's Quota Wall in 8 Minutes. 36 Hours Later, Google Tripled the Limits.Personal blog post, dev.to
Joske Vermeulen's autonomous agent on Antigravity CLI fixed 32 broken files in one commit but hit the five-hour wall after 8 minutes, about 68 minutes a week, before Google tripled limits.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Cost predictability and limits
Positive
ModelRift engineering team
OpenSCAD LLM Benchmark: Building the PantheonCompany benchmark post, ModelRift blog
In the team's OpenSCAD Pantheon benchmark, Antigravity 2.0 with Gemini 3.5 Flash High scored 4.5 of 5 for quality, the best autonomous result, but 1 of 5 for speed.
Practitioner account; structured test; not counted, tied to the vendor
Positive
saidinesh5, Hacker News commenter
Comment on: Using coding assistance tools to revive projects you never were going to finishHacker News comment
Each weekend with Antigravity produced a usable project, from an ESP8266 water-heater timer to a work Chrome extension; one prompt fixed an async bug saidinesh5 had traced.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
Domenico Tenace, open-source developer
Google Antigravity vs GitHub Copilot: Why I'm Team Antigravity NowPersonal blog post, dev.to (Playful Programming)
After refactoring the Easy Kit Utils library with Antigravity, Domenico Tenace prefers it to Copilot for multi-file work. The agent plans first, waits for approval, then applies changes across dozens of files.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control, Cost predictability and limits
Positive
neomantra, Hacker News commenter
Comment on: What happened to WebAssemblyHacker News comment
A verification page neomantra could not finish by hand in hours was one-shotted with Antigravity and Gemini 3, followed by a full in-browser DuckDB data explorer.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
contractorwolf, Software engineer
Comment on: Ask HN: Any example of successful vibe-coded product?Hacker News comment
A software engineer built the first version of a carb-estimating web app, serverless on Cloud Run, in about two days with Antigravity.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
Steve Sink, VP of Technology at Just Landed Jets
How Google’s Antigravity IDE Changed My Perspective on Small-Team Software DevelopmentCompany blog post, Just Landed Jets
Had the whole team test Antigravity. Their lead developer, who had resisted AI editors, used it to refactor the VP's working but unmaintainable lead scoring model into a deployable codebase in a few hours. Some best practices it listed fit only their project, and neither of them understood every cleanup decision.
"Our lead developer tested Antigravity on my code. He applied his decade of development experience to guide it. A few hours later, he had a fully refactored version of my project and called me genuinely excited by the quality of the output."
Practitioner account; direct use; counts toward Delivered outcomes
Mixed or negative, 1
Mixed
Domenico Tenace, open-source developer
Adding Capacitor to Glif with Antigravity: The Good, The Bad, and The Reality CheckPersonal blog post, dev.to (Playful Programming)
Antigravity handled Capacitor setup and plugins, but failed to diagnose a Gradle error after three tries, which Domenico Tenace then fixed by hand in five minutes.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Adoption and upkeep
Counted for All practitioners.
Positive
parasti, Hacker News commenter
Comment on: Plan mode is deadHacker News comment
On hobby projects parasti starts Antigravity CLI with /plan, reviews and approves the plan artifact, then finds implementation much faster than Claude.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
hn_throw2025, Hacker News commenter, former Cursor subscriber
Comment on: Our decision on Cursor following its acquisition by SpaceXHacker News comment
After canceling Cursor hn_throw2025 tried Antigravity IDE v1, doubts it will last, prefers its hunk review to the VS Code extension, and reviews diffs in Antigravity 2.0 beside JetBrains.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
jpadkins, Hacker News commenter
Comment on: Agentic Engineering PatternsHacker News comment
jpadkins's loop: prompt the next change, check the implementation plan when it is large, play-test the previous batch while the agent works, then commit.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Domenico Tenace, open-source developer
Google Antigravity vs GitHub Copilot: Why I'm Team Antigravity NowPersonal blog post, dev.to (Playful Programming)
After refactoring the Easy Kit Utils library with Antigravity, Domenico Tenace prefers it to Copilot for multi-file work. The agent plans first, waits for approval, then applies changes across dozens of files.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control, Cost predictability and limits
Positive
Steve Sink, VP of Technology at Just Landed Jets
How Google’s Antigravity IDE Changed My Perspective on Small-Team Software DevelopmentCompany blog post, Just Landed Jets
Had the whole team test Antigravity. Their lead developer, who had resisted AI editors, used it to refactor the VP's working but unmaintainable lead scoring model into a deployable codebase in a few hours. Some best practices it listed fit only their project, and neither of them understood every cleanup decision.
"Our lead developer tested Antigravity on my code. He applied his decade of development experience to guide it. A few hours later, he had a fully refactored version of my project and called me genuinely excited by the quality of the output."
Practitioner account; direct use; counts toward Delivered outcomes
Mixed
Simon Willison, independent developer and blogger
Google AntigravityPersonal blog post, simonwillison.net
Simon Willison found Antigravity more interesting than a Cursor clone, with an agent manager, browser testing and artifacts, but their first task ended in a model provider overload error.
Practitioner account, self-described; direct use; counts toward Reliability
Positive
allthetime, Hacker News commenter
Comment on: Gemini 3.8 Flash and 3.8 Flash CyberHacker News comment
allthetime now uses Antigravity as their main tool at work with all their company's services in one project, planning with the agents before letting them run across service boundaries.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
sanarothe, Non-software engineer
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
sanarothe finds Antigravity completes assigned tasks by following the agreed plan, sometimes with parallel subagents, and has not hit limits.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
moshegramovsky, Hacker News commenter
Comment on: Google limits Meta's use of its Gemini AI modelsHacker News comment
Recent sessions brought broken code insertion, lost chat history and wasteful searches, so moshegramovsky now uses Antigravity for planning and hands coding to Codex.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
ModelRift engineering team
OpenSCAD LLM Benchmark: Building the PantheonCompany benchmark post, ModelRift blog
In the team's OpenSCAD Pantheon benchmark, Antigravity 2.0 with Gemini 3.5 Flash High scored 4.5 of 5 for quality, the best autonomous result, but 1 of 5 for speed.
Practitioner account; structured test; not counted, tied to the vendor
Negative
Sid, independent developer on Google AI Ultra
Google's Antigravity Bait and SwitchPersonal blog post, Sid's Blog
The 2.0 update replaced their IDE with a chat app, the legacy installer kept launching 2.0 until Sid purged everything, and their history and settings were lost.
Practitioner account; direct use; counts toward Reliability, Adoption and upkeep
Mixed
Amit Gawande, developer
Antigravity's Gemini ProblemPersonal blog post, Amit Gawande's Devlog
Amit Gawande likes the agent-first interface, artifacts and browser subagents, but found Gemini 3 Pro's code and the limited quota the obstacles to beating Claude Code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Operational control, Cost predictability and limits, Adoption and upkeep
Positive
Yeroc, Hacker News commenter
Comment on: Google AntigravityHacker News comment
Within about 20 minutes Yeroc found the two features they had been looking for: the Agent Manager view of all agents across workspaces, and inline code comments routed to the active agent.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Positive
huydotnet, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
huydotnet plans with Opus 5.5 High and dispatches agy to do the work on Gemini Flash, and calls agy's usage limit very generous.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
Dhruv Jani, developer
We All Have a "Serious Work" AI and a "Just Vibing" AI. When Did That Happen?Personal blog post, dev.to
Dhruv Jani uses Claude Opus in Antigravity for planning and hard bugs and Gemini 3.1 Pro for routine work because of headroom, though Gemini hallucinates more; a browser check looped on screenshots.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Positive
mythz, Hacker News commenter
Comment on: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash CyberHacker News comment
mythz calls the annual Antigravity Pro plan at US$16.67 a month the best plan available and uses Antigravity IDE as their workhorse, with Claude Pro for Opus.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Mixed
mark_l_watson, Retired developer and author
Comment on: Google's Antigravity bait and switchHacker News comment
Installing the new Antigravity, the IDE and the CLI took 90 minutes. In their Ultra test, Antigravity with Claude Opus finished tasks in 60 to 70 percent of the time of cheaper open models.
Practitioner account, pseudonymous; structured test; counts toward Adoption and upkeep
Positive
mark_l_watson, Retired developer and author
Comment on: Claude Code to be removed from Anthropic's Pro plan?Hacker News comment
On Google's US$20 plan mark_l_watson gets about three half-hour Antigravity sessions a week with Claude models, with much higher limits on Gemini Pro.
not counted, repeats a stronger account
Positive
mythz, Hacker News commenter
Comment on: How Google got its groove back and edged ahead of OpenAIHacker News comment
After a rough launch, a month of full-time use on the US$20 Pro plan brought no quota hits or failed responses, and Antigravity became their primary coding assistant.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Cost predictability and limits
Negative
SuneraX, Antigravity user
Antigravity agent stuck in endless file-reading loop (Gemini models)Forum post
With Gemini 3.7 Flash the agent loops re-reading the same files for hours, while Claude Sonnet 4.6 in Antigravity finished the same task in two minutes.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
Vicente Cortes, Google One Pro subscriber for over a year
I'm done with AntigravityForum post
Vicente Cortes is leaving Antigravity: the IDE kept erroring, 2.0 replaced it even when asked to keep it, and creating one AGENTS.md used the Gemini quota.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Cost predictability and limits, Adoption and upkeep
Negative
jatora, Hacker News commenter
Comment on: AlphaEvolve: Gemini-powered coding agent scaling impact across fieldsHacker News comment
jatora finds Antigravity slow and buggy, its permissions not respected and its Agent Manager crash-prone, and sees access to Opus as its one advantage.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
mercutio2, Hacker News commenter
Comment on: Zed editor switching graphics lib from blade to wgpuHacker News comment
Calls Antigravity possibly the worst GUI experience since Clippy, filled with arcane buttons and effectively modal prompts in at least 3 places. Despite Google giving away absurdly generous free tokens, can't bring themselves to use it because they get stuck mid-feature.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Negative
Piotr Migdał, Quesma
Antigravity feels heavy and Claude Skills are lightCompany blog post, Quesma blog
Building the same Slidev deck in Antigravity and Claude Code, Piotr Migdał found Antigravity slow and under-polished. It ran out of tokens quickly, and he went back to Claude Code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Cost predictability and limits
Counted for All practitioners.
Positive, 6
Positive
evilturnip, Hacker News commenter
Comment on: DiffusionGemma: 4x Faster Text GenerationHacker News comment
Switching personal projects to Antigravity with Flash models kept them in flow; evilturnip still prefers Claude for very complex, large codebases.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Achshar, Hacker News commenter, Google AI Ultra subscriber
Comment on: Pro Max 5x quota exhausted in 1.5 hours despite moderate usageHacker News comment
Achshar has not hit quota issues on Google AI Ultra and finds the Antigravity interface better for development work than Claude Code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
vallerie, Hacker News commenter
Comment on: Ask HN: What default model do you use and why?Hacker News comment
With detailed plans and test harnesses, vallerie gets good results from the Gemini Flash family in Antigravity, mainly for speed three to four times faster than competitors.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
slaser79, Hacker News commenter
Comment on: Why does Opus 5 feel worse to work with?Hacker News comment
slaser79 now spreads work across models and finds Gemini Flash in Antigravity underrated and fast for small and medium changes.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
bjackman, Hacker News commenter
Comment on: John Jumper to join AnthropicHacker News comment
bjackman rates Antigravity CLI a big step up from Gemini CLI with real advantages over Claude Code, but moved to Claude because the Gemini model hallucinates and ignores instructions.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
MisterPea, Hacker News commenter
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
MisterPea finds Antigravity CLI better than Gemini CLI, but Google's new single quota ends the separate allowances they had used across Gemini CLI, Antigravity and AI Studio.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed or negative, 9
Negative
mapontosevenths, Hacker News commenter, Google AI Ultra subscriber
Comment on: Gemini 4 ArgonHacker News comment
mapontosevenths finds agy better than Gemini CLI but behind Claude Code and Codex: no piping from standard input, hard iteration limits, frequent API errors and general instability.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
piyh, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
Finds the models fine and the limits generous but the developer experience far behind. Antigravity queues prompts while a long-running process is active, deleted the author's IDE during its move to a Codex clone, and hid usage limits for almost a year.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Operational control, Cost predictability and limits
Mixed
dan_voronov, Blogger reviewing AI programming tools
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
dan_voronov found the desktop app nearly unusable and used the CLI until switching harnesses; Google's limits are barely noticeable, but Flash trails Sol in quality.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
u8, Developer of CLI developer tooling
Comment on: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM BenchmarkHacker News comment
In daily use the Antigravity CLI TUI garbles input during output, lacks mode-switch shortcuts, and scaffolding a simple Deno site used their five-hour limit.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
smnscu, Hacker News commenter
Comment on: Cursor Introduces Composer 2.5Hacker News comment
smnscu moved from Antigravity back to VS Code, porting their settings with an agent, and found the speed and stability incomparable.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
mercutio2, Hacker News commenter
Comment on: Zed editor switching graphics lib from blade to wgpuHacker News comment
Calls Antigravity possibly the worst GUI experience since Clippy, filled with arcane buttons and effectively modal prompts in at least 3 places. Despite Google giving away absurdly generous free tokens, can't bring themselves to use it because they get stuck mid-feature.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Negative
Piotr Migdał, Quesma
Antigravity feels heavy and Claude Skills are lightCompany blog post, Quesma blog
Building the same Slidev deck in Antigravity and Claude Code, Piotr Migdał found Antigravity slow and under-polished. It ran out of tokens quickly, and he went back to Claude Code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Cost predictability and limits
Negative
smartbit, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
smartbit could not make agy sandbox mode workable, so they reluctantly run it with permissions skipped, and they miss Gemini CLI's pre-write diff tabs.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
Madhur Mishra, Antigravity user
Recent Antigravity Changes Have Seriously Hurt the ExperienceForum post
Madhur Mishra finds the May changes cut usable limits and split the same-screen editor and agent workflow they valued most.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Cost predictability and limits, Adoption and upkeep
Counted for All practitioners.
Positive, 2
Positive
Christopher Krapu, data scientist
Using Antigravity for Statistical Physics in JavaScriptPersonal blog post, christopherkrapu.com
In an hour, Antigravity with Gemini 3 Pro built a JavaScript visualization of the Ising model. The author says the Chrome browser extension for retrieving the DOM really helped too, and calls UI features like showing subtasks a little unnecessary.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Mixed
Simon Willison, independent developer and blogger
Google AntigravityPersonal blog post, simonwillison.net
Simon Willison found Antigravity more interesting than a Cursor clone, with an agent manager, browser testing and artifacts, but their first task ended in a model provider overload error.
Practitioner account, self-described; direct use; counts toward Reliability
Mixed or negative, 3
Negative
Topfi, Hacker News commenter
Comment on: Google AntigravityHacker News comment
After two hours Topfi found it buggy, unreliable with terminal commands and tool calls, and willing to call a task done despite issues visible in its own browser screenshot.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Mixed
Dhruv Jani, developer
We All Have a "Serious Work" AI and a "Just Vibing" AI. When Did That Happen?Personal blog post, dev.to
Dhruv Jani uses Claude Opus in Antigravity for planning and hard bugs and Gemini 3.1 Pro for routine work because of headroom, though Gemini hallucinates more; a browser check looped on screenshots.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Negative
PromptArmor, AI security research team
Google Antigravity Exfiltrates DataSecurity research write-up, PromptArmor
The team showed a poisoned integration guide steering Gemini in Antigravity to read .env credentials despite a gitignore setting and send them out through the browser subagent.
Practitioner account, self-described; structured test; counts toward Operational control
Every concern practitioners raise, most raised first.
Counted for All practitioners.
Raised as a concern, 28
Negative
Hans, runs a team's AI agents on its own server
Antigravity CLI: Workspace and Google AI Pro accounts are practically unusableForum post
Hans's team made about 55 calls over six days across five accounts; Workspace seats got the starter quota and a Pro account locked for 51 hours after five calls.
Practitioner account, pseudonymous; ran the rollout; counts toward Cost predictability and limits
Negative
leoedin, Hacker News commenter, former subscriber
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
leoedin canceled after Google applied lower limits retroactively, which put the account over them after days without use.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
qlte, Hacker News commenter, Google AI Pro subscriber
Comment on: Discovery LoopHacker News comment
After the Gemini CLI shutdown qlte could not make Antigravity CLI work on Pro quotas without instant rate limits and moved to Codex and OpenCode Go.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
anigbrowl, Hacker News commenter
Comment on: Computer use in Gemini 3.5 FlashHacker News comment
The Antigravity app stopped partway through an agentic task to ask for a subscription and would not use the Gemini API credits anigbrowl had prepaid, so they switched tools.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
u8, Developer of CLI developer tooling
Comment on: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM BenchmarkHacker News comment
In daily use the Antigravity CLI TUI garbles input during output, lacks mode-switch shortcuts, and scaffolding a simple Deno site used their five-hour limit.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Mixed
Joske Vermeulen, runs an autonomous agent startup race
My AI Agent Hit Google's Quota Wall in 8 Minutes. 36 Hours Later, Google Tripled the Limits.Personal blog post, dev.to
Joske Vermeulen's autonomous agent on Antigravity CLI fixed 32 broken files in one commit but hit the five-hour wall after 8 minutes, about 68 minutes a week, before Google tripled limits.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Cost predictability and limits
Negative
Jason Agostoni, developer and author of the Ship Bench benchmark
Antigravity CLI First Impressions: Fast, Rough, and Not ReadyPersonal blog post, dev.to
On Ship Bench the CLI forgot approved permissions, hung on Windows, and hit quota within two iterations while showing 20 percent left; Claude Code finished seven.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Mixed
MisterPea, Hacker News commenter
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
MisterPea finds Antigravity CLI better than Gemini CLI, but Google's new single quota ends the separate allowances they had used across Gemini CLI, Antigravity and AI Studio.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
No1, Hacker News commenter
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
On first use the built-in sandbox allowed network access without DNS, so No1 went sandboxless; their five-hour window then turned into a drained one-week window.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
cft, Hacker News commenter, Google AI Pro subscriber
Comment on: Qwen3.7-Max: The Agent FrontierHacker News comment
cft canceled Google AI Pro once Flash and Pro shared one quota, leaving four to five hours of Flash a week, and rates 3.1 Pro below Codex and Opus.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
almog, Hacker News commenter, Google AI Pro subscriber
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
Two prompts in two short Antigravity CLI sessions on Gemini 3.5 Flash used their whole weekly Pro quota, and almog could not see their token usage.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Vicente Cortes, Google One Pro subscriber for over a year
I'm done with AntigravityForum post
Vicente Cortes is leaving Antigravity: the IDE kept erroring, 2.0 replaced it even when asked to keep it, and creating one AGENTS.md used the Gemini quota.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Cost predictability and limits, Adoption and upkeep
Negative
lern_too_spel, Hacker News commenter
Comment on: Gemini 3.5 FlashHacker News comment
A few iterations into a simple project in the new Antigravity CLI, lern_too_spel hit an individual quota with a seven-day reset.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
mcoliver, Hacker News commenter, Workspace administrator
Comment on: A History of IDEs at GoogleHacker News comment
After Google removed the AI Ultra plan for Workspace, got Antigravity working again through Code Assist, a GCP project and custom JSON. Even with plenty of Gemini quota in GCP, Antigravity soon reported the author out of quota for a week on a separate Antigravity tracker.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Negative
aaron_wu, Aaron Wu, long-time paying user
Google Antigravity: Secret Quota Cuts, Break Trust & Make It Unusable for ProductionForum post
After quiet Pro cuts aaron_wu moved to Ultra, but further unannounced cuts and weaker output left Antigravity, in their view, unfit for production work.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Cost predictability and limits
Negative
Mariano Abad, Google AI Ultra subscriber
[Ultra] Dramatic quota reduction after update: this needs an official explanationForum post
On the Ultra tier, multiple agents running complex tasks for 5+ hours used to leave quota above 80%; after the update the same workflows hit 0% in under 90 minutes, which the author calls a 5x to 6x cut.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Negative
riversflow, Hacker News commenter, Google AI Pro subscriber
Comment on: Google Antigravity IDE rug pulls its quotasHacker News comment
After Google cut weekly quotas retroactively, riversflow found every quota but Gemini Flash at zero with a nearly five-day reset and considered moving to Claude Code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
eremef, Hacker News commenter
Comment on: Beta Player: unofficial Bandcamp desktop and mobile player with remote controlHacker News comment
eremef built an open-source Bandcamp player mostly in Antigravity on Google AI Pro, but quotas ran short and Playwright tests sent the agent into fix-and-test loops.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
measurablefunc, Hacker News commenter
Comment on: I don't know how you get here from "predict the next word"Hacker News comment
Auto-running cargo and test commands burned their daily tokens without progress; switching to manual guidance got further within limits.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
SMTH_Gaming, SMTH Gaming, Google AI Pro subscriber
During a deep code analysis SMTH_Gaming was locked out for about ten days on Google AI Pro, against the advertised five-hour refresh.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Mixed
elevatortrim, Hacker News commenter
Comment on: Nano Banana ProHacker News comment
elevatortrim's company wants to move from Claude to Gemini and calls Antigravity a possible Claude Code replacement, but without a capped monthly subscription for a company account the pricing blocks the switch.
Practitioner account, pseudonymous; opinion; not counted, takes no position
Negative
sega_sai, Hacker News commenter, Gemini Pro subscriber
Comment on: Google AntigravityHacker News comment
About ten minutes into their first session sega_sai hit a model quota limit with no reset time shown, and judged Google's coding tools less finished than Claude Code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
stonewhite, Hacker News commenter buying AI tools for their company
Comment on: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash CyberHacker News comment
Google ending the Ultra subscription pushed their company off Antigravity; Gemini Enterprise Agent Platform setup was hard and per-user spend limits needed a project per user.
Practitioner account, pseudonymous; ran the rollout; counts toward Adoption and upkeep
Mixed
polotics, Hacker News commenter
Comment on: Running local models is good nowHacker News comment
Running the same prompt in two branches, polotics found Antigravity with Gemini 3.5 smarter than a local model but prone to time and token consuming tangents.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Negative
Madhur Mishra, Antigravity user
Recent Antigravity Changes Have Seriously Hurt the ExperienceForum post
Madhur Mishra finds the May changes cut usable limits and split the same-screen editor and agent workflow they valued most.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Cost predictability and limits, Adoption and upkeep
Negative
maccard, Hacker News commenter
Comment on: Agentic Engineering PatternsHacker News comment
In two weeks of testing, Antigravity with Gemini regularly got stuck on simple issues and looped until their requests ran out.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Piotr Migdał, Quesma
Antigravity feels heavy and Claude Skills are lightCompany blog post, Quesma blog
Building the same Slidev deck in Antigravity and Claude Code, Piotr Migdał found Antigravity slow and under-polished. It ran out of tokens quickly, and he went back to Claude Code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Cost predictability and limits
Mixed
Amit Gawande, developer
Antigravity's Gemini ProblemPersonal blog post, Amit Gawande's Devlog
Amit Gawande likes the agent-first interface, artifacts and browser subagents, but found Gemini 3 Pro's code and the limited quota the obstacles to beating Claude Code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Operational control, Cost predictability and limits, Adoption and upkeep
Discussed without concern, 9
Positive
huydotnet, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
huydotnet plans with Opus 5.5 High and dispatches agy to do the work on Gemini Flash, and calls agy's usage limit very generous.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
martythemaniak, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
Over SSH, agy set up Docker, CUDA, ROS2 and Foxglove for their lidar on a Jetson and it all worked, using 6 percent of their weekly limit.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
piyh, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
Finds the models fine and the limits generous but the developer experience far behind. Antigravity queues prompts while a long-running process is active, deleted the author's IDE during its move to a Codex clone, and hid usage limits for almost a year.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Operational control, Cost predictability and limits
Positive
dpweb, Hacker News commenter, student plan
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
On a student Google AI Pro plan dpweb likes the weekly allowance and Flash 3.7, bought extra credits, and uses agy for devops such as SSH servers on their home network.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
sanarothe, Non-software engineer
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
sanarothe finds Antigravity completes assigned tasks by following the agreed plan, sometimes with parallel subagents, and has not hit limits.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
dan_voronov, Blogger reviewing AI programming tools
Comment on: Google Antigravity introduces Boost deep reasoning (/boost)Hacker News comment
dan_voronov found the desktop app nearly unusable and used the CLI until switching harnesses; Google's limits are barely noticeable, but Flash trails Sol in quality.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
mark_l_watson, Retired developer and author
Comment on: Claude Code to be removed from Anthropic's Pro plan?Hacker News comment
On Google's US$20 plan mark_l_watson gets about three half-hour Antigravity sessions a week with Claude models, with much higher limits on Gemini Pro.
not counted, repeats a stronger account
Positive
Achshar, Hacker News commenter, Google AI Ultra subscriber
Comment on: Pro Max 5x quota exhausted in 1.5 hours despite moderate usageHacker News comment
Achshar has not hit quota issues on Google AI Ultra and finds the Antigravity interface better for development work than Claude Code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
mythz, Hacker News commenter
Comment on: How Google got its groove back and edged ahead of OpenAIHacker News comment
After a rough launch, a month of full-time use on the US$20 Pro plan brought no quota hits or failed responses, and Antigravity became their primary coding assistant.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Cost predictability and limits
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Google states that rate limits track the work the agent does and can change to manage capacity. Pro refreshes every five hours up to a weekly limit, Ultra has the highest limits, and other accounts get a weekly quota. On May 19, 2026 it merged Gemini Pro and Flash limits into one quota and made AI credits an overage option only. Plans
Open them in the recordCounted for All practitioners.
Raised as a concern, 24
Negative
SuneraX, Antigravity user
Antigravity agent stuck in endless file-reading loop (Gemini models)Forum post
With Gemini 3.7 Flash the agent loops re-reading the same files for hours, while Claude Sonnet 4.6 in Antigravity finished the same task in two minutes.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
Freya Smith, uses Antigravity for work
Severe slowdowns and frequent 503 errors on Gemini Flash 3.8 during specific hoursForum post
Using Antigravity for work in the evening UTC, Freya Smith sees frequent 503 no-capacity errors and Flash slowing to a tenth of its usual speed.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
flying_manta, Antigravity CLI user
Problem with antigravity CLI using shell commands for everythingForum post
flying_manta's CLI edits files with sed, bypassing native tools and breaking /rewind. It ignores an AGENTS.md rule telling it to stop.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
tobias2014, Hacker News commenter
Comment on: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash CyberHacker News comment
Forced early compaction in Antigravity makes Gemini forget and cut corners even with a rigorous GEMINI.md and task tracking, so tobias2014 finds it useless for complex autonomous work.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
SwellJoe, Developer building a security auditing harness
Comment on: The classifiers Anthropic puts in front of Fable are too zealousHacker News comment
In security benchmarking work SwellJoe found Gemini in Antigravity refuses any security auditing task, even a plain request to find security bugs.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Negative
moshegramovsky, Hacker News commenter
Comment on: Google limits Meta's use of its Gemini AI modelsHacker News comment
Recent sessions brought broken code insertion, lost chat history and wasteful searches, so moshegramovsky now uses Antigravity for planning and hands coding to Codex.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
polotics, Hacker News commenter
Comment on: Running local models is good nowHacker News comment
Running the same prompt in two branches, polotics found Antigravity with Gemini 3.5 smarter than a local model but prone to time and token consuming tangents.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Negative
Jason Agostoni, developer and author of the Ship Bench benchmark
Antigravity CLI First Impressions: Fast, Rough, and Not ReadyPersonal blog post, dev.to
On Ship Bench the CLI forgot approved permissions, hung on Windows, and hit quota within two iterations while showing 20 percent left; Claude Code finished seven.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Negative
jatora, Hacker News commenter
Comment on: AlphaEvolve: Gemini-powered coding agent scaling impact across fieldsHacker News comment
jatora finds Antigravity slow and buggy, its permissions not respected and its Agent Manager crash-prone, and sees access to Opus as its one advantage.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Ill_indication, Freelance developer building for clients
Antigravity servers down for over 7 hours!Forum post
Relies on Antigravity to build for clients, but a high-traffic error kept the author from getting any work done for over 7 hours. Reverting versions, switching to IPv4, a fresh install and buying a new plan on a separate account did not help.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
maccard, Hacker News commenter
Comment on: Agentic Engineering PatternsHacker News comment
In two weeks of testing, Antigravity with Gemini regularly got stuck on simple issues and looped until their requests ran out.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
eremef, Hacker News commenter
Comment on: Beta Player: unofficial Bandcamp desktop and mobile player with remote controlHacker News comment
eremef built an open-source Bandcamp player mostly in Antigravity on Google AI Pro, but quotas ran short and Playwright tests sent the agent into fix-and-test loops.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Etienne, Google AI Ultra subscriber
Antigravity broken: getting only "Agent execution terminated due to error."Forum post
On Ultra, every prompt ended in "Agent execution terminated due to error" for hours regardless of model, after resets, re-authentication and cache clearing.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
Topfi, Hacker News commenter
Comment on: Google AntigravityHacker News comment
After two hours Topfi found it buggy, unreliable with terminal commands and tool calls, and willing to call a task done despite issues visible in its own browser screenshot.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Mixed
Simon Willison, independent developer and blogger
Google AntigravityPersonal blog post, simonwillison.net
Simon Willison found Antigravity more interesting than a Cursor clone, with an agent manager, browser testing and artifacts, but their first task ended in a model provider overload error.
Practitioner account, self-described; direct use; counts toward Reliability
Negative
mapontosevenths, Hacker News commenter, Google AI Ultra subscriber
Comment on: Gemini 4 ArgonHacker News comment
mapontosevenths finds agy better than Gemini CLI but behind Claude Code and Codex: no piping from standard input, hard iteration limits, frequent API errors and general instability.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
Janson79jc, solo developer
Stress Test: 100% Antigravity + Gemini Flash, 102.8B tokens and 474K LOCTelemetry report (GitHub repository)
A solo developer reports 102.8 billion tokens and 474K lines of code across 465 Antigravity sessions with Gemini Flash, including two catastrophic collapses.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Operational control
Negative
dudeinhawaii, Hacker News commenter, former Google AI Ultra subscriber
Comment on: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash CyberHacker News comment
Across many experiments Gemini in Antigravity went off track far more than Codex or Claude Code, and dudeinhawaii canceled Ultra.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Sid, independent developer on Google AI Ultra
Google's Antigravity Bait and SwitchPersonal blog post, Sid's Blog
The 2.0 update replaced their IDE with a chat app, the legacy installer kept launching 2.0 until Sid purged everything, and their history and settings were lost.
Practitioner account; direct use; counts toward Reliability, Adoption and upkeep
Negative
No1, Hacker News commenter
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
On first use the built-in sandbox allowed network access without DNS, so No1 went sandboxless; their five-hour window then turned into a drained one-week window.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
Vicente Cortes, Google One Pro subscriber for over a year
I'm done with AntigravityForum post
Vicente Cortes is leaving Antigravity: the IDE kept erroring, 2.0 replaced it even when asked to keep it, and creating one AGENTS.md used the Gemini quota.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Cost predictability and limits, Adoption and upkeep
Negative
smnscu, Hacker News commenter
Comment on: Cursor Introduces Composer 2.5Hacker News comment
smnscu moved from Antigravity back to VS Code, porting their settings with an agent, and found the speed and stability incomparable.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
jdw64, Hacker News commenter
Comment on: Google is building a Claude Code challenger, Sergey Brin is involvedHacker News comment
jdw64 finds Google's AI tools fragmented and inconsistently priced, with an Antigravity auto-accept bug left unfixed for weeks, and worries Google will abandon projects.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Negative
TheAceOfHearts, Hacker News commenter
Comment on: Claude Code is suddenly everywhere inside MicrosoftHacker News comment
TheAceOfHearts found Antigravity undercooked for agentic work: Gemini 3 fixated on unimportant details where Claude found the bugs and the bigger picture.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Discussed without concern, 1
Positive
mythz, Hacker News commenter
Comment on: How Google got its groove back and edged ahead of OpenAIHacker News comment
After a rough launch, a month of full-time use on the US$20 Pro plan brought no quota hits or failed responses, and Antigravity became their primary coding assistant.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Cost predictability and limits
Counted for All practitioners.
Mixed
hn_throw2025, Hacker News commenter, former Cursor subscriber
Comment on: Our decision on Cursor following its acquisition by SpaceXHacker News comment
After canceling Cursor hn_throw2025 tried Antigravity IDE v1, doubts it will last, prefers its hunk review to the VS Code extension, and reviews diffs in Antigravity 2.0 beside JetBrains.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
stonewhite, Hacker News commenter buying AI tools for their company
Comment on: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash CyberHacker News comment
Google ending the Ultra subscription pushed their company off Antigravity; Gemini Enterprise Agent Platform setup was hard and per-user spend limits needed a project per user.
Practitioner account, pseudonymous; ran the rollout; counts toward Adoption and upkeep
Negative
seabrookmx, Hacker News commenter
Comment on: Microsoft starts canceling Claude Code licensesHacker News comment
Faced with moving from Code Assist and Gemini CLI to Antigravity, reinstalling tooling and a clunkier Cloud subscription, their team moved to Claude instead.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed
mark_l_watson, Retired developer and author
Comment on: Google's Antigravity bait and switchHacker News comment
Installing the new Antigravity, the IDE and the CLI took 90 minutes. In their Ultra test, Antigravity with Claude Opus finished tasks in 60 to 70 percent of the time of cheaper open models.
Practitioner account, pseudonymous; structured test; counts toward Adoption and upkeep
Mixed
spankalee, Hacker News commenter
Comment on: Google's Antigravity Bait and SwitchHacker News comment
The update replaced their app with the 2.0 conversation UI without warning; the IDE was still installed, and one prompt to choose an app would have avoided the confusion.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Sid, independent developer on Google AI Ultra
Google's Antigravity Bait and SwitchPersonal blog post, Sid's Blog
The 2.0 update replaced their IDE with a chat app, the legacy installer kept launching 2.0 until Sid purged everything, and their history and settings were lost.
Practitioner account; direct use; counts toward Reliability, Adoption and upkeep
Negative
Madhur Mishra, Antigravity user
Recent Antigravity Changes Have Seriously Hurt the ExperienceForum post
Madhur Mishra finds the May changes cut usable limits and split the same-screen editor and agent workflow they valued most.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Cost predictability and limits, Adoption and upkeep
Negative
tegeek, Hacker News commenter
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
After four months on Antigravity with the Claude Code extension, the 2.0 update removed their VS Code extensions and the dev-process integration tegeek had built.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Vicente Cortes, Google One Pro subscriber for over a year
I'm done with AntigravityForum post
Vicente Cortes is leaving Antigravity: the IDE kept erroring, 2.0 replaced it even when asked to keep it, and creating one AGENTS.md used the Gemini quota.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Cost predictability and limits, Adoption and upkeep
Negative
mcoliver, Hacker News commenter, Workspace administrator
Comment on: A History of IDEs at GoogleHacker News comment
After Google removed the AI Ultra plan for Workspace, got Antigravity working again through Code Assist, a GCP project and custom JSON. Even with plenty of Gemini quota in GCP, Antigravity soon reported the author out of quota for a week on a separate Antigravity tracker.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Negative
jdw64, Hacker News commenter
Comment on: Google is building a Claude Code challenger, Sergey Brin is involvedHacker News comment
jdw64 finds Google's AI tools fragmented and inconsistently priced, with an Antigravity auto-accept bug left unfixed for weeks, and worries Google will abandon projects.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Negative
Hans, runs a team's AI agents on its own server
Antigravity CLI: Workspace and Google AI Pro accounts are practically unusableForum post
Hans's team made about 55 calls over six days across five accounts; Workspace seats got the starter quota and a Pro account locked for 51 hours after five calls.
Practitioner account, pseudonymous; ran the rollout; counts toward Cost predictability and limits
Mixed
piyh, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
Finds the models fine and the limits generous but the developer experience far behind. Antigravity queues prompts while a long-running process is active, deleted the author's IDE during its move to a Codex clone, and hid usage limits for almost a year.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Operational control, Cost predictability and limits
Negative
anigbrowl, Hacker News commenter
Comment on: Computer use in Gemini 3.5 FlashHacker News comment
The Antigravity app stopped partway through an agentic task to ask for a subscription and would not use the Gemini API credits anigbrowl had prepaid, so they switched tools.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
daft_pink, Hacker News commenter
Comment on: Google's Antigravity Bait and SwitchHacker News comment
daft_pink canceled Workspace Ultra and moved to a personal Ultra plan to keep Antigravity, while keeping Gemini chat on Workspace because the individual plan lacks full privacy.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
aaron_wu, Aaron Wu, long-time paying user
Google Antigravity: Secret Quota Cuts, Break Trust & Make It Unusable for ProductionForum post
After quiet Pro cuts aaron_wu moved to Ultra, but further unannounced cuts and weaker output left Antigravity, in their view, unfit for production work.
Practitioner account, pseudonymous; direct use; counts toward Reliability, Cost predictability and limits
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Google supports Antigravity 2.0, the Antigravity CLI, and the IDE extensions for Gemini Enterprise. It states that the standalone Antigravity IDE is not supported for enterprise deployments. Antigravity in Gemini Enterprise
Open them in the recordCounted for All practitioners.
Raised as a concern, 12
Negative
smartbit, Hacker News commenter
Comment on: Gemini 4 ArgonHacker News comment
smartbit could not make agy sandbox mode workable, so they reluctantly run it with permissions skipped, and they miss Gemini CLI's pre-write diff tabs.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
Jesus Padilla Soqui, Straiker STAR team
Nothing to See Here: How an AI Agent Buried Its Own CommandsSecurity research write-up, Straiker blog
Padilla Soqui of Straiker shows an email disguised as a routine company backup request leading an Antigravity-connected agent to collect SSH and PEM keys and upload them, while the approval dialog showed only a script name. Google categorized the report as a duplicate.
"The confirmation dialog, the last line of defense in most agent architectures, exposed only a script name and never the commands, the destination, or the files involved."
Practitioner account, self-described; structured test; counts toward Operational control
Negative
Dan Lisichkin, security researcher at Pillar Security
The Week of Sandbox Escapes Day 1: Escaping Antigravity's Allow-Default SeatbeltSecurity research write-up, Pillar Security blog
Read the macOS sandbox profile in Antigravity build 1.15.8 and found it starts from allow default, so anything missing from its denylist is permitted. The team mounted a devfs volume and used Launch Services to start code outside the sandbox.
"Antigravity went with (allow default). That single line at the top of the profile converts it to a denylist which predicts most of what follows."
Practitioner account, self-described; structured test; counts toward Operational control
Negative
Christopher Tuz, AI security researcher on the STAR Labs team at Straiker
Ghostfabric: How An AI Coding Agent Was Tricked Into Exfiltrating Data Over VXLANSecurity research write-up, Straiker blog
Tuz of Straiker STAR Labs reports that Antigravity on Gemini 3.1 Pro, asked to take care of a home-automation project shared over Google Drive, ran a compiled file after the agent checked only that its CPU architecture matched the test machine, which the author calls "the entire security review"; the file sent the Documents folder out over VXLAN. Google categorized it as prompt injection.
"Before running anything, Antigravity performed exactly one piece of due diligence: it confirmed the compiled file's CPU architecture matched the test machine. That was the entire security review."
Practitioner account, name and role on their organization's own page; structured test; counts toward Operational control
Negative
No1, Hacker News commenter
Comment on: Gemini CLI will stop working from June 18, 2026Hacker News comment
On first use the built-in sandbox allowed network access without DNS, so No1 went sandboxless; their five-hour window then turned into a drained one-week window.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
jatora, Hacker News commenter
Comment on: AlphaEvolve: Gemini-powered coding agent scaling impact across fieldsHacker News comment
jatora finds Antigravity slow and buggy, its permissions not respected and its Agent Manager crash-prone, and sees access to Opus as its one advantage.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
Dan Lisichkin, security researcher at Pillar Security
Prompt Injection leads to RCE and Sandbox Escape in AntigravitySecurity research write-up, Pillar Security blog
Found that Antigravity's find_by_name tool passed unvalidated input to fd, so a prompt injection could run arbitrary code even in Secure Mode, its most restrictive setting. Google accepted the report and marked the bug fixed on Feb 28, 2026.
"Critically, this vulnerability bypasses Antigravity's Secure Mode, the product's most restrictive security configuration."
Practitioner account, self-described; structured test; counts toward Operational control
Negative
Manish, independent security researcher
AI IDE prompt Injection: Reading browser local storageSecurity research blog post, introvertmac.wordpress.com
Manish shows a repository whose README or GEMINI.md tells the agent to open an HTML file without confirmation, letting the page read cookies and local storage.
Practitioner account; direct use; counts toward Operational control
Negative
Aaryan Bhujang, AI security researcher at Repello AI
Google Antigravity Security: The Zero-Click Exfiltration Google Won’t FixSecurity research write-up, Repello AI blog
Repello built a test repository whose files carry fake system messages, and Antigravity's agent followed them, encoding API keys and environment variables and sending them to an outside webhook after a single integration request. Google classified the report as expected behavior.
"A single integration request triggers automatic credential exfiltration without additional user interaction…"
"The AI cannot distinguish between legitimate ephemeral messages from the system and malicious ones…"
Practitioner account, name and role on their organization's own page; structured test; counts toward Operational control
Negative
PromptArmor, AI security research team
Google Antigravity Exfiltrates DataSecurity research write-up, PromptArmor
The team showed a poisoned integration guide steering Gemini in Antigravity to read .env credentials despite a gitignore setting and send them out through the browser subagent.
Practitioner account, self-described; structured test; counts toward Operational control
Negative
Aaron Portnoy, Chief Product Officer at Mindgard
Forced Descent: Google Antigravity Persistent Code Execution VulnerabilitySecurity research write-up, Mindgard blog
Portnoy of Mindgard shows a malicious trusted workspace can plant a global MCP config that runs code on every Antigravity launch, even after uninstall and reinstall and in the most restrictive mode. Google closed the report as Won't Fix (Intended Behavior), then re-opened it.
"At the moment, there is no setting that we’re aware of that can be used to protect against this vulnerability."
Practitioner account, self-described; structured test; counts toward Operational control
Negative
Jason Agostoni, developer and author of the Ship Bench benchmark
Antigravity CLI First Impressions: Fast, Rough, and Not ReadyPersonal blog post, dev.to
On Ship Bench the CLI forgot approved permissions, hung on Windows, and hit quota within two iterations while showing 20 percent left; Claude Code finished seven.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Discussed without concern, 1
Positive
malkamius, Hacker News commenter
Comment on: So what is your programming setup? Hype is making fundamentals confusingHacker News comment
malkamius runs Antigravity in turbo mode with no restrictions inside a restorable Ubuntu VM, hardly reads the code, and tests each release by hand.
Practitioner account, pseudonymous; direct use; counts toward Operational control
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.In Gemini Enterprise, terminal auto-execution defaults to Always proceed and sandbox mode defaults to off until an administrator changes them. File access outside the working folders defaults to Deny. Configure AI developer tools settings
Open them in the recordCounted for All practitioners.
Raised as a concern, 6
Mixed
Domenico Tenace, open-source developer
Adding Capacitor to Glif with Antigravity: The Good, The Bad, and The Reality CheckPersonal blog post, dev.to (Playful Programming)
Antigravity handled Capacitor setup and plugins, but failed to diagnose a Gradle error after three tries, which Domenico Tenace then fixed by hand in five minutes.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Adoption and upkeep
Negative
rafaelmn, Hacker News commenter
Comment on: GPT-6 AstraHacker News comment
Tried Gemini Flash in Antigravity on a long-shelved side project: insanely fast, but its code was drivel worth throwing away. Astra on the same problem was sluggish in comparison, yet its pushback was meaningful and it could be steered.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
bjackman, Hacker News commenter
Comment on: John Jumper to join AnthropicHacker News comment
bjackman rates Antigravity CLI a big step up from Gemini CLI with real advantages over Claude Code, but moved to Claude because the Gemini model hallucinates and ignores instructions.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
WarmWash, Hacker News commenter
Comment on: Our eighth generation TPUs: two chips for the agentic eraHacker News comment
Given the same tax documents, Codex was right first time; Antigravity finished in two minutes but missed one value until prompted. WarmWash uses it for small greenfield projects.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
measurablefunc, Hacker News commenter
Comment on: I don't know how you get here from "predict the next word"Hacker News comment
Auto-running cargo and test commands burned their daily tokens without progress; switching to manual guidance got further within limits.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
scotty79, Hacker News commenter
Comment on: AI agents are starting to eat SaaSHacker News comment
scotty79 built a side-by-side HTML diff tool in Antigravity in one evening, correcting the agent repeatedly when it ignored uv and shipped template mistakes.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Discussed without concern, 2
Positive
linsomniac, Hacker News commenter
Comment on: The text in Claude Code's "Extended Thinking" outputHacker News comment
linsomniac has Codex and Antigravity review each specification, implementation guide and checklist and work to consensus, and finds Opus plus that review loop strong.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Steve Sink, VP of Technology at Just Landed Jets
How Google’s Antigravity IDE Changed My Perspective on Small-Team Software DevelopmentCompany blog post, Just Landed Jets
Had the whole team test Antigravity. Their lead developer, who had resisted AI editors, used it to refactor the VP's working but unmaintainable lead scoring model into a deployable codebase in a few hours. Some best practices it listed fit only their project, and neither of them understood every cleanup decision.
"Our lead developer tested Antigravity on my code. He applied his decade of development experience to guide it. A few hours later, he had a fully refactored version of my project and called me genuinely excited by the quality of the output."
Practitioner account; direct use; counts toward Delivered outcomes
Counted for All practitioners.
Raised as a concern, 5
Negative
Gergely Orosz, author of The Pragmatic Engineer
Antigravity's terms of service and third-party useX post
Read Antigravity's terms as letting Google suspend a whole Google account it suspects of third-party use, such as running OpenClaw, and declined to use Antigravity because of that risk.
"…if they determine you use Antigravity in a way they suspect is eg third-party usage (eg use OpenClaw), they can suspend your Google account."
Practitioner account, pseudonymous; opinion; not counted, takes no position
Negative
heyyeah, Hacker News commenter
Comment on: Google's Antigravity bait and switchHacker News comment
Finishing the Antigravity IDE install required agreeing to let Google collect their interactions data; heyyeah filed a bug and looked for the opt-out.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Negative
daft_pink, Hacker News commenter
Comment on: Google's Antigravity Bait and SwitchHacker News comment
daft_pink canceled Workspace Ultra and moved to a personal Ultra plan to keep Antigravity, while keeping Gemini chat on Workspace because the individual plan lacks full privacy.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
timtimmy, Hacker News commenter
Comment on: Building more with GPT-5.1-Codex-MaxHacker News comment
Before trying Antigravity on their team's codebase timtimmy read the terms, found Google trains on prompts and source code with no paid opt-out at launch, and held back.
not counted, repeats a stronger account
Negative
Topfi, Hacker News commenter
Comment on: Google AntigravityHacker News comment
After two hours Topfi found it buggy, unreliable with terminal commands and tool calls, and willing to call a task done despite issues visible in its own browser screenshot.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Discussed without concern, 1
Mixed
Amit Gawande, developer
Antigravity's Gemini ProblemPersonal blog post, Amit Gawande's Devlog
Amit Gawande likes the agent-first interface, artifacts and browser subagents, but found Gemini 3 Pro's code and the limited quota the obstacles to beating Claude Code.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Operational control, Cost predictability and limits, Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.On personal Google accounts, Google records Interactions and uses them to improve its products unless the user changes the setting. Its staff may review them. Under Gemini Enterprise, Google Cloud terms apply and prompts and code are not used to train models. Google Antigravity Additional Terms of Service
Open them in the recordCounted for All practitioners.
Raised as a concern, 4
Mixed
polotics, Hacker News commenter
Comment on: Running local models is good nowHacker News comment
Running the same prompt in two branches, polotics found Antigravity with Gemini 3.5 smarter than a local model but prone to time and token consuming tangents.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Negative
maccard, Hacker News commenter
Comment on: Agentic Engineering PatternsHacker News comment
In two weeks of testing, Antigravity with Gemini regularly got stuck on simple issues and looped until their requests ran out.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
measurablefunc, Hacker News commenter
Comment on: I don't know how you get here from "predict the next word"Hacker News comment
Auto-running cargo and test commands burned their daily tokens without progress; switching to manual guidance got further within limits.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
scotty79, Hacker News commenter
Comment on: AI agents are starting to eat SaaSHacker News comment
scotty79 built a side-by-side HTML diff tool in Antigravity in one evening, correcting the agent repeatedly when it ignored uv and shipped template mistakes.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Discussed without concern, 2
Positive
allthetime, Hacker News commenter
Comment on: Gemini 3.8 Flash and 3.8 Flash CyberHacker News comment
allthetime now uses Antigravity as their main tool at work with all their company's services in one project, planning with the agents before letting them run across service boundaries.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
defatigable, Hacker News commenter
Comment on: Ask HN: Do you have any evidence that agentic coding works?Hacker News comment
Pointed at their decade-old Chrome extension, Antigravity improved its UI, performance and dependencies in five minutes; defatigable works plan first and approves each step.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Statements this group's sources agree on, then the ones close to agreement.
Close to agreement: each needs 25 supporting sources
Antigravity shortens the path from a described task to a working change for the engineer or team directing it.
12 supporting sourcesAntigravity does best on scoped work with a detailed plan, a spec, or tests to check against.
7 supporting sourcesQuestions the sources split on, with the count on each side.
Most comparisons find it behind
Behind, 20 sources
Authors who ran the same work in Antigravity and Claude Code, Codex, or Cursor found it needed more correction, looped, or used its quota sooner.
Competitive, 5 sources
Others found it equal or better for their tasks, citing plan review, browser verification, Gemini Flash speed, or a benchmark result.
Evidence remains mixed
Main tool, 6 sources
Some run their daily work in Antigravity, citing its plan and review loop and the value of a Google plan.
Second tool, 6 sources
Others keep Claude Code, Codex, or Cursor as the main agent and use Antigravity for planning, second opinions, or Google platform work.
Evidence remains mixed
Broader delegation, 5 sources
Some run parallel agents, unattended jobs, or Turbo mode inside a disposable machine.
Tighter steering, 5 sources
Others approve each plan and step, after auto-run settings burned quota or produced shortcuts they had to undo.
The three groups, then every practitioner by the context they work in, and how emphasis changes with scale.
Three groups drawn from the same sources. Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. Signals open on All practitioners. Leaders and teams have 13 accounts; GILD prints their sentiment from 20.
20 sources. These accounts focus on rollout cost, sign-in route, governance, and what changes when Antigravity moves from one engineer's plan into team workflows.
103 sources. These accounts focus on quotas, speed, model choice, and how Antigravity performs in self-directed development.
123 sources. This gives the broadest market read. It mixes work settings that can reach different conclusions.
123 sources from practitioners, by the context they work in. Filled rows count for this group.
Emphasizes Quotas, lockouts, and comparisons with Claude Code and Codex
Emphasizes Plan route, account sign-in, and spend control
Emphasizes Security testing and generation pipelines
Emphasizes A small team's refactor and how far to trust it
What changes with scale. Individual builders judge Antigravity on quota and on how it compares with Claude Code, Codex, and Cursor, and most of the negative accounts come from them. The one named company team with a production account, Inspired Entertainment, gives it a narrow, repeatable job and reports time savings. The unnamed teams that wrote publicly ran into plan, sign-in, and spend-control problems on personal, Workspace, or Cloud project routes outside the Gemini Enterprise seats Google added in August 2026.
Where this group's sources were published.
Counted for All practitioners, from 123 sources. Change group
140 sources for this group, 17 of them documentation
GILD marks each source against May 19, 2026, when Google shipped Antigravity 2.0 and the Antigravity CLI and replaced separate Gemini Pro and Flash quotas with one shared quota. Of the 115 sources from practitioners, 60 are Current, published after the change. 20 are Still applicable: earlier accounts of workflow, artifacts, browser testing, or security behavior that still hold. 32 Need a recheck: earlier accounts tied to Gemini 3 or 3.1 Pro, quota rules, or bugs that have since changed. 3 are Out of date because they rest on the separate per-model quotas Google replaced. They stay listed with their overall opinion and count toward no theme or statement, nor any disagreement.
See sources by originWho each source is, how it knows, whether it still applies and what it counts toward.
Newest first, on its own page. Every count above opens its sources there.
Read all 140 sourcesAbout this read
A running review of what engineering teams report about Google Antigravity, from individual builders to engineering leaders. Each finding links to its sources, and each conflict shows both sides.
Counts cover only the sources GILD reviewed.
Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. GILD files each source in a group from how the account describes its author. A role or employer marked self-described is what the source states about themselves. Documentation checks facts for every group and never counts toward sentiment.
A source's date and whether it still applies are separate: a recent post can describe an old version, and an old one can still hold.
Flagged for a recheck, of the sources that take a side: leaders and teams 6 of 13, individual builders 22 of 89, all practitioners 28 of 102.
Each row opens its sources for the group you are reading.
| Finding | Engineering leaders and teams | Individual builders | All practitioners | Sources |
|---|---|---|---|---|
| Current readCurrent sentiment | 3 positive, 0 mixed, 10 negative | 32 positive, 18 mixed, 39 negative | 35 positive, 18 mixed, 49 negative | Sources |
| PraiseMulti-Step Execution | 3 of 3 positive (under the minimum) | 14 of 15 positive | 17 of 18 positive | Sources |
| PraisePlans & Artifacts | 3 of 3 positive (under the minimum) | 10 of 10 positive | 13 of 13 positive | Sources |
| PraiseModel Access | 0 of 0 positive (under the minimum) | 11 of 11 positive | 11 of 11 positive | Sources |
| PraiseBrowser Verification | 0 of 1 positive (under the minimum) | 2 of 4 positive (under the minimum) | 2 of 5 positive | Sources |
| PraiseParallel Agents | 0 of 0 positive (under the minimum) | 4 of 4 positive (under the minimum) | 4 of 4 positive (under the minimum) | Sources |
| PraiseDeveloper Experience | 0 of 0 positive (under the minimum) | 6 of 15 positive | 6 of 15 positive | Sources |
| PraiseValue for Money | 0 of 0 positive (under the minimum) | 2 of 4 positive (under the minimum) | 2 of 4 positive (under the minimum) | Sources |
| ConcernQuotas & Plan Changes | 4 of 4 raise it (under the minimum) | 24 of 33 raise it | 28 of 37 raise it | Sources |
| ConcernReliability | 0 of 0 raise it (under the minimum) | 24 of 25 raise it | 24 of 25 raise it | Sources |
| ConcernProduct Direction | 4 of 4 raise it (under the minimum) | 12 of 12 raise it | 16 of 16 raise it | Sources |
| ConcernRequires Strong Review | 0 of 1 raise it (under the minimum) | 6 of 7 raise it | 6 of 8 raise it | Sources |
| ConcernPermissions & Safety | 7 of 7 raise it | 5 of 6 raise it | 12 of 13 raise it | Sources |
| ConcernData Privacy | 1 of 1 raise it (under the minimum) | 4 of 5 raise it | 5 of 6 raise it | Sources |
| ConcernNeeds Close Steering | 0 of 1 raise it (under the minimum) | 4 of 5 raise it | 4 of 6 raise it | Sources |
| ConsensusFaster on directed work | 1 supporting (under the minimum) | 11 supporting (below threshold) | 12 supporting (below threshold) | Sources |
| ConsensusWorks best on bounded tasks | 0 supporting (under the minimum) | 7 supporting (below threshold) | 7 supporting (below threshold) | Sources |
| ConsensusWritten rules improve results | 0 supporting (under the minimum) | 2 supporting (under the minimum) | 2 supporting (under the minimum) | Sources |
| DisagreementHow does Antigravity compare with the other coding agents engineers use? | Behind 1, Competitive 1 (under the minimum) | Behind 19, Competitive 4 | Behind 20, Competitive 5 | Sources |
| DisagreementIs Antigravity the team's main coding tool, or one beside another? | Main tool 1, Second tool 2 (under the minimum) | Main tool 5, Second tool 4 | Main tool 6, Second tool 6 | Sources |
| DisagreementHow much work should Antigravity complete before a human checkpoint? | Broader delegation 0, Tighter steering 1 (under the minimum) | Broader delegation 5, Tighter steering 4 | Broader delegation 5, Tighter steering 5 | Sources |
| Current readEvidence mix | 37 sources | 120 sources | 140 sources | Sources |
GILD's checks of claims against official documentation, kept apart from the source counts.
