Gmail Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

Gmail Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Grannybull (@Grannybull69) reported 3 minutes ago

  @SolisCode @ProtonSupport Yea. Me, too. I switched from @CenturyLink e-mail as it was down for weeks, so not sure where to go. gmail maybe?

• 
  Frank Ricard (@NovusOrderM) reported 3 minutes ago

  Our firm has been in the process of building our own software, almost entirely coded by AI, that is unique to the needs of our firm. As this has been developed, utilization in Microsoft Windows is going down. Not just Excel, Word, or PowerPoint, but even file storage is migrating into the system itself. For 30+ years, the deal was you build your digital enterprise on top of Windows, whether you use 20% of it or 80% of it. There really wasn't an alternative. Generic software for everyone, because custom software was too expensive and time-consuming to build. AI is breaking that constraint. Whether you're a manufacturer, a bank, an investment firm, etc. Now each sector, industry, and individual enterprise can build their own software, cheap and fast, to do exactly what the company needs it to do and nothing more. First thought: Amazing and a level of incredulity. Second thought: Fine, there's still going to be a communication layer that survives. Common language/software between enterprises to share information and communicate - email, spreadsheets, PDFs, etc. So Microsoft et al keep the interop layer even as it loses the workspace. But what happens when AI agents are handling company-to-company communications? Why would they need Outlook, Gmail, Excel, etc., at all? They don't need a common file format designed for humans. They'll negotiate their own mode of communication or build one. So the endgame might actually be: custom software inside every company, agent-to-agent communications between companies, and the universal platform layer - the thing that has anchored enterprise computing since the 90s - just... no longer necessary. Not sure on the timeline, but watching what is happening at my own company feels less like a prediction and more like a lagging indicator.

• 
  Magma (@master_ronix) reported 3 minutes ago

  @VinayakGavariya Yup regeneration button. Zoho connectors are available but we can't sign in indus android app using arattai or zohomail address. Only option available are phone number & gmail. Finally about stop button, when quota ends we can't push new prompt to same chat or session i guess.

• 
  Adi Leviim | Building SaaS Products (@adileviim) reported 5 minutes ago

  we built auto tagging that sorts every conversation into one of five buckets: coding, writing, research, math and science, business. no custom tags, no way to add a sixth category. that was a deliberate constraint, not a v1 shortcut we meant to fix later. the first version we prototyped let people create their own tags, unlimited, freeform, like gmail labels. it demoed great and died in real use. within two weeks test users had eleven overlapping tags for the same handful of topics, 'work stuff', 'job', 'career', all meaning roughly the same thing to the same person on different days. a tag system with infinite categories quietly becomes a tag system with zero categories, because nothing gets filed the same way twice. five fixed buckets forced consistency we couldn't get any other way. a coding conversation is always coding, no matter what mood the user was in when they had it. that consistency is the entire value of automatic tagging, if the same kind of chat can land in a different bucket depending on nothing, the feature isn't saving anyone time, it's just adding a layer of noise between them and search. the tradeoff is real and we don't pretend it isn't. someone doing legal research or medical writing gets folded into research or writing whether they like the label or not, and a few people have asked for narrower categories. we said no, on purpose, because the moment you add a sixth bucket you're on the road to fifty, and the whole point was to stop that road before it starts. the lesson generalizes past tagging. any time you're building automatic classification, the number of categories matters more than how smart the classifier is. a great model sorting into too many buckets still produces a mess a human has to clean up. a dumb model sorting into a small, well chosen set produces something people actually trust enough to stop thinking about.

• 
  Nick (@nickbaumann_) reported 6 minutes ago

  On Friday night I had two Uber Eats orders from the same restaurant cancelled back-to-back, each after waiting ~45 minutes I was refunded for both, but frontline support refused to give me anything for the two wasted hours I told ChatGPT Work not to stop until it got me some kind of compensation, then let it handle the rest 15 minutes of active work and ~2.5 hours of background follow-up later, it got me $95.46 in Uber Cash here’s the full sequence: 6:11 PM: first order cancelled after a ~45 minute wait. I was refunded, but got nothing for the wasted time 6:19 PM: I told ChatGPT Work to get me some kind of consolation and not back down. It opened a support case and asked for $20 in Uber Cash 6:22 PM: the first agent refused any credit or promo. ChatGPT rejected the resolution and asked for a supervisor, then tried a promo, fee waiver, Uber One extension, or membership adjustment instead 6:31 PM: a second agent also refused compensation. ChatGPT kept the case open and got it transferred to the Uber One team 7:18 PM: my second order was also cancelled after another ~45 minute wait. I was refunded again, but still received no compensation. I told ChatGPT to keep going 7:25 PM: ChatGPT opened a separate Uber One case covering both failures, raised the ask to $30, and pushed the agent to transfer it to a specialist 7:33 PM: ChatGPT submitted a separate written complaint with both cancellation times, the second order ID, and my annual Uber One membership details 7:34 PM: ChatGPT confirmed my Uber Wallet was still at $0 and set up background checks across both support cases, Gmail, and the wallet 7:50 PM: an Uber specialist approved $40.62 for the first cancellation and $54.84 for the second, totaling $95.46 in Uber Cash 8:44 PM: ChatGPT found the response, verified the credit in my actual Uber Wallet, marked the goal complete, and shut down its own follow-up automation credit where it’s due: this was a genuinely frustrating experience, but Uber ultimately made it right (& then some tbh) and treated me like a longtime, frequent customer

• 
  abir (@abiryanii) reported 10 minutes ago

  @mattpocockuk @BenjDicken @PlanetScale We do this but for tool quality regarding common errors (agent uses Gmail search filters incorrectly, misuses input args, etc) and it compounds really well! Some of our toolkits have over 100 versions over their lifetime if you take a look

• 
  Henrique Meyer (@HenriqueMeyerFL) reported 12 minutes ago

  @poteto @bot I signed up for the Supergrok Heavy last Thursday. Grok Bot went through an entire week's worth of usage on day 1, and another $1000+ in on-demand usage since then. I love the concept, but this is ridiculous. Grok Bot is incredible, but here are some major problems I'm having during my first week. 1. Since all the bots share a computer, they are all logged into the same Chrome sessions. Fine -- but I have different projects with different accounts that a bot needs to be logged into for each one. So I created Chrome profiles. The profiles seem to get wiped clean frequently, so I had to make a bot to police the other bots and force them to stay in the correct profiles for their projects. The plugins also only connect to one account each. Like Gmail only connects to one Gmail account. 2. I can chat with regular Grok all day and not reach any kind of token limit. If I try to work through the same things with Grok Bot, it burns through so money so quickly. But I want the bots to have the context of the conversation I'm having, and using regular Grok is completely separate from anything I'm working on in Grok Bot. 3. One bot, one screen. Why can't the same bot do things on multiple screens at the same time? I'm limited to one computer at a time because I'm human and I only have 1 set of eyes and 2 hands. I shouldn't have to spin up several bots for something that should be covered under the same role, just so different tasks can be done asynchronously. One bot brain with multiple screens should be able to handle tasks a lot faster, and skip having to communicate between different bots to get something done. 4. Authentication/Passkey/WindowsHello/2FA always requires human interaction. Any way to set up a password store and Passkeys on the Bot computer to avoid me having to re-login all the time? 5. Bot looks like it's "thinking" but the screen is still. What's going on? Am I getting billed for this silent time? Need way more transparency regarding how much usage is happening, or I'll go bankrupt before Grok Bot actually has a chance to help bring in revenue.

• 
  jx (ultraviolet) (@jx) reported 19 minutes ago

  Ok, I'm just done with @google now. Search is almost like roulette now, I have no idea what kind of results I'll get back. It's garbage. I know they're forcing people to use Gemini now, but even that won't answer the questions without a fight. Maybe I'll go look at Kagi. I finally permanently moved off of Chrome a while ago. I'm working on leaving GMail, too. I still pay for YouTube Premium because it's great, but that's it. It's the only Google property left that actually does what I want. The sad thing is, I'm only one of millions, and @google doesn't care. By the time they finally try to fix search it'll be too late, the momentum will have turned. They underestimate the power of momentum. AI has made writing code easy and completely ruined search.

• 
  Amna Aslam (@amnaaslam20) reported 21 minutes ago

  @NCbassey Just log out of all your Gmail accounts. After logging off, go to LinkedIn sign in, and then just add your Gmail account. LinkedIn just adds our running Gmail account when we try to log in.

• 
  SIR ALEXANDER (@SirAlexanderrr) reported 21 minutes ago

  That means emails sent to either your old or new Gmail address will appear in the same inbox. You can also sign in to Google services using either address.

• 
  lilAgents (@lilAgents) reported 21 minutes ago

  To be clear, renting a tool isn't the problem. Gmail is a rental and that's fine, because you can export your data and leave anytime. The problem is a platform like Wix or Squarespace, where there's no export path and a price hike leaves you stuck. That's what ownership is really protecting against: your exit.

• 
  Architect🛡️ (@Architect9000) reported 22 minutes ago

  MCP design is hugely underappeciated. Optimizing both the shape and implementation of tools can yield incredible cost and performance improvements. Treating MCP as a thin API wrapper is a terrible mental model. Codependent, optional, or too many tool parameters lead to false negatives. Poor server and tool descriptions thwart the agent's tool selection reasoning step. Lack of normalization transfers deterministic work on the API side into nondeterministic work for the agent. High quality error responses are nearly as important as the success case. I recently experimented with tool response Large Result Offloading and saw token consumption drop by nearly 30% in data-rich applications. For agentic workloads, input tokens are like throwing a pebble in a pond. The original token cost is multiplied by cacheability and compaction concerns later. Just a little more thoughtfulness in MCP design can dramatically improve tool success rate and token efficiency, especially for complex surfaces. I could design a Gmail MCP that would drop costs by 50% versus a generic wrapper. And since the tool layer is a common abstraction point, the gains compound across an entire organization.

• 
  Paul Walsh (@Paul__Walsh) reported 28 minutes ago

  I’ve just spent hours responding to an avalanche of comments about the ChatGPT plugin for Apple Messages. Here are the main ones: 💬 "Anyone can screenshot your messages" Yes. They can also print them and fax them to the FBI. People using end to end encryption accept the risk that another participant can deliberately betray their trust. They don’t expect that person to give third party software ongoing access to read, search and analyse their private conversations. 💬 "I opted in, so it’s not a problem" You opted in. Everyone you communicate with didn’t. 💬 "The encryption still works, so you’re wrong" Yes, the cryptography still works. My concern is third party access to the readable content before or after decryption. 💬 "The recipient could always share your messages" Yes. The recipient could betray your trust. That doesn’t justify automating third party access to your private conversations. 💬 "Any Mac app with Full Disk Access could already do this" Yes. Apple specifically warns that Full Disk Access can expose Messages. OpenAI has now productised that capability for ordinary people who aren't aware of the potential implications, at scale. I don't think Apple saw this coming. 💬 "This is no different from giving AI access to email, SMS, Slack or Teams" It is. Those services don’t provide the same end to end encryption model as iMessage. Think Signal, not SMS. This Proton, not Gmail. The screenshot one is my favourite. I’ll either update my original article with an FAQ or write a dedicated follow up. Thankfully, these comments represent less than 2% of the responses I’ve received here. I haven’t even looked at LinkedIn yet. My concern with this implementation is scale. One person can enable it on their Mac and expose accessible conversations involving potentially hundreds of people who never enabled ChatGPT, never agreed to the access and aren’t notified. Depending on their Apple and iCloud configuration, that can include historical conversations, participant identifiers, contact information and attachments. Proton summarised my argument well after reading the article. I recommend reading their post first, then clicking through to mine for the full analysis. My Substack is free. I’m building a community around corporate surveillance, government overreach, privacy and personal digital safety. I’m interested in what the technology can do, what information it can access, who else can be affected and what happens when that capability is deployed at scale.

• 
  Theo - t3.gg (@theo) reported 28 minutes ago

  I think we’ll quickly hit some kind of balance here where software goes into a few buckets: 1. AI-powered human interfaces (Codex, T3 Code, Claude Cowork) 2. programmatic interfaces (anything with a good MCP, CLI, or agent ready APIs) 3. legacy human interfaces (apps like Excel, Gmail, Photoshop) Investor world seems to think everything will become category 2. Anyone who’s tried to use one of these awful bridges knows that will never happen. I can’t even automate my GitHub usage without hitting endless rate limits 🙃 Category 1 is getting saturated fast with chat interfaces. Category 3 is fighting to pretend they still matter by baking awful agents into their products. I think Category 2 is the best place to fight right now, but you have to do it without pretending you can just staple a slow, rate limited API onto some existing software. I think @pierrecomputer is the best example right now. It seems like fixing GitHub to handle the unique needs and load from agents isn’t viable, and they built something tailor made for the needs of agents.

• 
  Marcus (@biggerd05) reported 29 minutes ago

  @Timmy2Tap14 @Google @gmail Did you ever find a fix? Still not receiving notifications.