I came back to Codex in Visual Studio Code. It works slower, but after a few prompts, all my bugs are gone, and the brainstorming capacity is so superior to what I got with Antigravity. I’m confused. Eventually, I will also check Claude.
2 Likes
I also want to add my support to the devs here and express my frustration. I personally subscribed to the Ultra plan exclusively for Claude Opus 4.6, and I sometimes find myself with absolutely no quota left after just 4 prompts.
Moreover, I’ve never really liked Gemini for coding, but over the last few days, I feel like Gemini 3.1 Pro has completely regressed (which is ironic since I code for Android, so you’d expect Google to have an edge). I don’t know if I’m being objective here, but even for extremely simple tasks without heavy context, it’s become a total disaster. I’m constantly having to fix its mistakes, repeat myself, and confront it about its hallucinations, even when using excellent workflows.
Anyway, I haven’t had the time to read through everything here; I just stumbled onto this forum to make sure I wasn’t an isolated case, and it’s pretty clear that I’m not.
4 Likes
You are certainly not alone 
I moved to VSCode with Claude 2 days ago and its lightyears ahead.
3 Likes
i loved antigravity, it was my main IDE about three months, after these quota updates i rearly use, it is ridiculuous this quota rates, you need to support your software google, that is how you get attention from users. i am really sad about this.
3 Likes
I only send 3 meesage then … it consume 40% off my credit …
1 Like
With this type of response on forum and to my ticket. I am really not sure Pramanik you are a person or just AI response.
2 Likes
Are there, in truth, any pertinent colleagues affording genuine consideration to this matter?
1 Like
I opened my antigravity after 6 to 7 days and after giving antigravity my first prompt with Gemini (low), I checked my quota and its showing 60% remaining and 2 days cooldown timer. How is this possible that my quota reduce even when I haven’t use Antigravity for last 4-5 days.
I expect quotas to apply for the duration of my 1-hear subscription at least, though. They’re literally not providing what I they said I was paying for. If it’s not criminal it’s certainly unethical.
1 Like
Hello all,
In my opinion, Google’s ecosystem has grown too heavy and sluggish. Relying on their services for long-term projects feels increasingly risky.
The support is practically non-existent. There’s no help on the Antigravity official site, and their social media presence is dead. It took a lot of digging just to find this community, and even here, activity is minimal (shoutout to @Abhijit_Pramanik for the lone reply recently).
Just as Google once disrupted the old guard, I believe it’s time to shift our focus to newer, more agile tools.
Between Codex, Cursor, and Claude, which do you think is the best investment for a developer right now? Any suggestions?
4 Likes
Completely agree. The silence, the hidden quota reductions, and the drop in quality are destroying trust. It’s very sad to see Google treat paying users like this. In the long run, behavior like this will cost them far more than they think, because trust is what decides who wins in AI.
3 Likes
i’ve been using Kimi via their official vscode extension for actual work, the 40$ plan seems to be a pretty good deal, but also i’m SUPER cautious doing stuff now since i have PTSD from the antigravity horrors. i also keep the 10$ monthly Trae on standby when i need to get that zerocontextbias 2nd opinion on a quick fix so my workflow is uninterrupted. i typically avoid direct api since the usage is so low, doing coding its local, and its nice to see these companies giving us IDE usage, however many are tending to give us those “inflated” api type pricings.
can really go with any of them that you vibe with, its not a big committment. could essentially sample 5 different ones for 20 bucks each and have your own personal feedback for less than half the cost of ultra experiments…
2 Likes
I have been using Claude Max ($200) through VSCode for the past few days. I am coding for 13 hours straight on a very complex project and it hasn’t even touched 30% capacity and I’m running Opus4.6 On MAX Effort.. It can run for 2 hours working on complex refactors, planning and coding and its quality is miles better than what antigravity was. This is a no brainer. I would say most people will get away with the 90 dollar plan if you are a hobbyist etc.
GUYS, DON’T WASTE ANYMORE TIME ON THIS CHANNEL. Just move to Claude.
Over and out..
6 Likes
Just picked up the Claude Code Pro plan today for testing. It’s actually mind-blowing: in my experience, the ‘Ultra’ quota on Antigravity is roughly equivalent to what you get on a basic $20 Claude Pro sub. Absolute madness. Heck of a job, Google
3 Likes
Is this actually true? That would be insane.
Its completely true. Complete rip off from Google.
I have Claude max 200, been flat out doing coding, refactoring and Playwright testing for 6 hours today, barely scratching the quota.. I haven’t had to stop and made back the 3 weeks lost productivity from this debacle by Google… I even got a free 187 dollars of over use credits I can’t see myself touching this month.
with AntiG.. was getting 20 mins on ultra.
don’t waste your time here. get a refund from Google in your subscription and get your Claude direct.
the model is on FULL MAX and 1 million context on Opus 4.6 and my code base is very large and complex application. Quality so much better and no limit on MCP services…
3 Likes
@Manorama_Namboori You locked my post for what good reason - say something you seem to sure be active enough and have some type of opinion but you don’t say anything, deleting my posts every chance you get - whats your problem. Getting kinda sick of your stalking and harassment.
2 Likes
An upgrade is supposed to signify an increase in efficiency. However, Google’s latest update has done the exact opposite, transforming a once-fluid development experience into a grueling nightmare. First, the technical design flaws are disheartening; since the update, search functionality within the Terminal frequently freezes, forcing workflows to a grinding halt. Even worse, the previously efficient “Always Allow” permission mechanism has been gutted, forcing developers to manually confirm operations within the Workspace constantly. This deliberate increase in friction is essentially pushing developers toward third-party plugins—a move that constitutes technical self-sabotage for a platform claiming to be “all-in-one.”
Compounding this is the extreme instability of the infrastructure. Despite Google owning the world’s largest server clusters, the current user experience is riddled with “Server Busy” alerts and unannounced session cutoffs. This instability, combined with the aforementioned permission and search bugs, creates a malicious “Token Drain” trap. Every time a tool hangs or a server crashes, developers are forced to restart sessions, re-describe problems, and re-load context. This has effectively doubled the rate of token consumption. Even for those holding both Pro and Ultra accounts, the tokens disappear instantly within this inefficient loop. We are paying premium fees only to foot the bill for the system’s own failures.
From a product strategy perspective, there is a cynical, almost intentional feel to this decline. Compared to the meticulous craftsmanship seen in the developer experiences of Cursor or Claude, Google seems to be sacrificing core intelligence for the sake of monetization. There is a palpable “nerfing” of Gemini 3.1 Pro’s reasoning capabilities. This appears to be a form of architectural hegemony. Google is mandating that the other model(claude, gpt) route tasks through its internally developed—yet profoundly inefficient—Antigravity search and analysis tools. This design strips the model of its native high-level logic and traps the developer in a “tool-calling trap” characterized by frequent hangs and errors. This cycle of inefficiency failing to retrieve files, halving debugging speed, and shrinking intelligence, which has resulted in a productivity level that is not even half of what it was in previous versions.
As a global leader in both software and hardware, Google’s current product quality is an utter disqualification when measured against competitors like Cursor and Claude. This is no longer just a technical issue; it is a display of total disregard for the needs of professional developers by a tech giant chasing tool-based monetization. We don’t need more “useful tools” layered onto a broken core. We need a stable, intelligent, and respectful underlying architecture that values the developer’s time and resources. Unless this vicious cycle is broken, these so-called upgrades will continue to alienate and disillusion Google’s most loyal professional users.
6 Likes
I’m still going. Its 20.15 here (GMT). Literally non stop.. Claude Max.. halft the price of Google AntiG and about x1000 productivity..
3 Likes
It was supposed that Google would be more generous… Given the fact that their model is also behind the competitors in terms of quality output. Quite disappointed.
1 Like