Hey everyone, I think it’s time we have a serious, honest conversation about the current state of the Google AI ecosystem. Ever since early April, and especially following the recent I/O 2026 rollout, the platform’s stability hasn’t just dipped—it has completely nosedived.
Right now, Gemini feels like the most unstable major AI model on the market. As power users and developers, we are constantly dealing with the dreaded “infinite loading loop” where the model just gets stuck on “Thinking…” forever and completely fails to output anything.
It’s not just the UI freezing up. The backend management is a mess. We are seeing aggressive, silent quota downgrades where we get shoved onto the weaker 3.1 Flash Lite without any transparent error messages, breaking our complex workflows. On top of that, the new silent safety filters are literally erasing full responses right before they render on our screens, leaving the UI hanging in limbo.
And it’s not just Gemini. NotebookLM is currently in shambles. The recent push for resource-heavy Studio features seems to have crippled the servers. Trying to load a simple PDF often results in a permanent loading screen or a blank white page, which the Workspace Status Dashboard has officially confirmed as an ongoing disruption.
I know the push for the “Agentic Era” with 3.5 Flash, Gemini Spark, and Omni is exciting for Google. But as the people actually using and building on this platform, we can’t rely on a foundation that crumbles under its own weight every few hours. We don’t need faster benchmark specs on paper; we need a system that actually works reliably in reality.
Please hit the brakes on the feature bloat and prioritize infrastructure stability.
You are not the only one. It has been degrading over time and suddenly it feels like a chatbot from the past. I’m just concerned if this is the future of the other apps too
Spot on. The regression is exactly what is so alarming. It really does feel like a massive step backward, especially when the new Gemini 3.5 Flash still struggles with outdated knowledge cutoffs that literally make it act like a chatbot from the past.
Your concern about the rest of the ecosystem is completely valid. We are already seeing this degradation bleed into the regular Google Assistant, not to mention the ongoing stability catastrophe and critical regressions happening with NotebookLM right now. It feels like they are sacrificing the reliability of core, functioning tools just to force this new ‘agentic’ infrastructure on us before the backend is actually ready to handle it.
I feel you. I use Gemini for my professional rendering work. It had its limitations, but it always got the job done.
As of now, June 2026, Gemini is an expensive DISASTER. Does not recognize my prompts anymore, and is widely inconsistent. I am looking for valid alternatives, but It’s tough, since all AI models have absurd monthly prices.
I guess Google shifted its attention to the new “Gemini Spark”
Glad I’m not the only one. Gemini lately has been an absolute desaster for me. Especially in tasks where it absolutely excelled before.
It doesn’t complete tasks. Sometimes it says it’s calculating, but instead just stops
It’s absolute half-assed - full of errors, glitches and unresponsiveness
It does not matter if you switch to the pro model - sometimes pro is even WORSE than flash
For example I tasked it with a fairly simple Google Sheets setup - it utterly failed. Calculated for 5 Minutes and created a file that had one line of text in the first cell of the sheet. Absolute waste of time and ressources.
Sometimes it imagines that it did a task, but didn’t execute it. It’s like it’s hallucinating to itself.
It tells you “okay I will fix it” and just stops. Only if I switch to an aggressive tone it finally does something - which is wrong again. I hope they fix it.
I’m stunned at the interface breakages alone. The “+” button now trying to do everything (options and modes), but visibly loading in two stages, reminds me of the “adaptive” menus trend in old software. I don’t understand how no one seems to be in charge of the actual UI design side of things. I had Gemini Pro trial for three months, and it has put me off paying for Gemini for at least another year. To be fair, I also get multiple fault report emails daily from ChatGPT, and I find Claude hard to use, but Google is big enough to really win at this AI game, so why don’t they focus properly on making it work? Although Apple have so many faults, perhaps the Apple Siri deal will make Google realise the value of a proper interface on stable software. It’s 2026, by now, software for public use should just work. For fun, give this sort of webpage to Gemini and watch it bite the hand that makes it.
Your experience with the Google Sheets setup failing and the model “hallucinating” that it finished the task is actually a documented issue right now known as the “False Completion State” bug. And the reason Pro sometimes feels worse than Flash is due to their aggressive new routing system. If the servers hang or you hit a hidden limit, the system stealth-downgrades your prompt to the much weaker 3.1 Flash Lite behind the scenes without even telling you. It’s an absolute waste of time and resources.
I couldn’t agree more about wanting the old version back. The new safety filters are completely broken. Instead of just blocking a prompt up front and giving you a warning, the system now generates the response internally and then silently erases the text right before it renders on your screen. It leaves the UI hanging in a loop, which is exactly why it feels like the model just gave up and “stopped” working.
You nailed it on the UI breakages. That “+” button loading in stages is a perfect example of how disjointed the frontend has become. It feels like there’s zero quality control on the design side right now because they are frantically scrambling to force these new “agentic” workflows onto a backend that clearly can’t handle the traffic. Your point about the Apple/Siri deal is spot on—it’s 2026, and if the basic interface can’t even load a menu properly, what are we even paying a premium subscription for?
Unfortunately they’ll never do that. They keep chasing the biggest goose to compete with the other AI platforms instead of prioritizing stability and reliability. Back when 2.5 PRO was the latest model, most people praised Gemini and encouraged others to use it instead of the other platforms because it was the best. Now? I’ve seen most people talk about Claude or even GPT, Gemini being the last resort option. The only way for any changes to happen is if their revenue take a big hit, which means most people cancelling their subscription and moving elsewhere. Unfortunately, that will never happen. As long as they make money, they got no reason to make any meaningful changes.
Man, you hit the nail right on the head. It’s a tough pill to swallow, but you’re completely right. They’re so obsessed with the AI arms race and pushing out these shiny new “agentic” features to stay in the headlines that they’ve entirely neglected the foundation we all rely on.
I miss the 2.5 Pro days too. Back then, the platform was rock solid, and I used to recommend it to everyone over ChatGPT. Now? It’s honestly hard to defend Gemini when Claude is sitting right there, working flawlessly without getting stuck in infinite thinking loops or silently erasing its own answers right in front of us.
You’re spot on about the revenue hit being the only real wake-up call. It’s incredibly frustrating because so many of us are tied deeply into the Google ecosystem, making it hard to just jump ship overnight. But you’re right—voting with our wallets is the only language corporate actually speaks. Until those paid subscription numbers take a massive dive, we’re basically just paying a premium to beta-test their broken updates.
About to cancel my paid subs as the products are excessively broken. Not just infinite loading and thinking forever, but repeating the same response to a previous query over and over and ignoring all other inputs. Really disappointed. And no desktop app… It’s bad.