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:

• 
  Leon Ni (@aileadworld) reported 3 minutes ago

  I heard early this morning that Andrew Ng released an open-source local agent. I was really curious—a project from the "Father of Google Brain" bound to have something unique, so I rushed to look it up. Sure enough, Andrew Ng posted a thread introducing their newly released project, OpenWorker. I’ve translated his post for everyone below: To summarize briefly: OpenWorker's goal is simply to get work done—organizing files, generating reports, replying on Slack, updating calendars, processing inboxes, checking task progress, and more. Put even simpler, it’s designed to deliver finished work straight to the user. But I have to ask: Is this really the open-source product built by the multi-titled "Father of Google Brain"? Project Overview OpenWorker is a local agent application open-sourced by Andrew Ng. It can connect to local files, the terminal, Slack, GitHub, Gmail, Google Calendar, Notion, and other tools. All you need to do is give it a goal. It will then break down the steps, call the necessary tools, ask for required approvals, and ultimately deliver documents, spreadsheets, reports, message replies, or other work outputs. DEMO Above is the DEMO released by Andrew Ng's team; below is mine. This time, I truly understood what "the higher the expectations, the greater the disappointment" means. I review open-source projects all the time, and usually, I check out each new repo with a calm, neutral mindset. Today, knowing Andrew Ng released something, I was genuinely excited. I rushed to test it out to see a master at work. Let’s start with the pre-compiled installer version. Great. It spent the entire morning loading, and the model still failed to initialize. Since I couldn't modify the .dmg file, I had no choice but to deploy it from source code. It lacks multi-language support, so for the sake of demonstrating it clearly here, I added a quick localization tweak myself. Next, I configured the LLM provider—no major issues here; most standard options are covered, though some netizens complained that OpenRouter support is missing. Once the task started, the "interesting" parts began—starting with the bug that triggered the most backlash in the comments section. First issue: I just wanted to run a quick test, so I typed in a prompt: "Help me summarize the current state of Agents." Since I was using a Chinese input method (IME), typing the English word "Agent" required pressing Enter to confirm the text selection. The moment I hit Enter, half of my unfinished prompt got sent out automatically. That said, its mechanism of clarifying requirements through iterative questions is actually something I quite like—the clearer the requirement, the more accurate the output. It also includes various integrations (connectors) to link tools like Slack—which, honestly, isn't much different from connecting OpenClaw to Lark/Feishu. Final Thoughts I rarely criticize an open-source project. No matter the motivation, anyone contributing to the open-source community deserves credit. If I don't like a project, I usually just choose not to recommend it. But honestly, I fail to see the value in Andrew Ng releasing this project. How is this any different from OpenClaw??? The comment section is filled with the exact same question. Is this really a product that should come from a legendary figure in the AI space? I’m not saying every release needs to be ground-breaking, but at least include a few novel ideas. How is this different from putting a new UI wrapper on OpenClaw? Even if there’s no innovation, could you at least test it before releasing it to fix basic bugs? It’s not that I’m holding you to unreasonable standards—it’s that you should safeguard your own reputation and deliver a product that matches your stature. Gathering thousands of stars right after launch means you should be responsible to the people who supported you. Seriously, I beg you: please do something more meaningful! Use AI for healthcare to benefit humanity, or focus on education to teach and inspire... I sincerely hope top-tier AI leaders think this through.

• 
  Ruth Bush (@RuthBush444) reported 5 minutes ago

  Yeah, just keep them coming, and I’l keep blocking them. No more car ads. Too bad they have to lock down my Google account to forward ads for cars, for the entire neighborhood….what will it be next? First it was my Gmail, then iCloud, then Facebook. TELL ME PLEASE.

• 
   James ★ ★ ★ ★ ☆™ (@jame_solution) reported 6 minutes ago

  We help creators regain access to their channels and fix account-related issues worldwide 💯 Channel not accessible YouTube login problems Hacked or compromised channel Lost Gmail access Brand Account issues Get support for your YouTube account problem Inbox 👇👇 @jame_solution

• 
  Arafa Abdul (@arafa_abdu67911) reported 7 minutes ago

  Good day, me using Google Authenticator to login and my Gmail account has been hacked. How am my going to sign in to my Sidra wallet account?

• 
  pupupu (@0xpupupu) reported 7 minutes ago

  OPUS 5 JUST DROPPED. IT WORKS MORE AUTONOMOUSLY THAN EVER Nobody sets guardrails after the damage. They set them 30 seconds before connecting the tool, or they don't set them at all. Opus 5 launched this week built to run longer and think for itself before asking u anything. It verifies its own work and recovers from errors without stepping in for permission first. That's power. It's also more steps where it can touch something u never meant to give it access to. 1. Draft-only, always. AI never sends anything on its own, Gmail, Slack, whatever. It writes the draft, a human hits send. 1 rule, 0 exceptions. 2. No delete permission, period. Not files, not emails, not records. If a tool doesn't let u separate "write" from "delete," that's ur first red flag before u even connect it. 3. New tool, new guardrails. Don't reuse the same rules for Gmail and for a payment API. Different tool, different blast radius. 4. Ask the AI to write its own limits. Before connecting anything unfamiliar, prompt it: "walk me through what hard limits and guardrails I should consider for this specific tool." It knows its own risk surface better than most people guessing. 5. Confirm before anything irreversible runs. Sends, deletes, payments, force-pushes. If it can't be undone, it needs a human eyeball first, no exceptions for "trusted" agents. The agent isn't careless. The setup just never asked it to be careful. What's the 1 guardrail u wish u'd set up before it was too late?

• 
  Rseghier (@rseghier) reported 7 minutes ago

  @james406 Try the Gemini feature in paid Gmail version. It’s a bit slow, but it’s very good at finding specific emails (or at extracting data from emails)

• 
  Roopa Srikanth (@RoopaNataraj) reported 9 minutes ago

  Anyone facing issues with Gmail today?! Unable to get my folders/drafts etc - the side navigation panel

• 
  Rambo (@monster_zero123) reported 10 minutes ago

  @timeout_7 @Anmolkakkar27 No. Use desktop mode on chrome. Use a dummy gmail acct to register/login.

• 
  Monster Mankar (@MonsterMankar) reported 12 minutes ago

  @GoogleIndia I am unable to login to my gmail account even my gpay is not working.. need help

• 
  FSD Life (@Tatakae360) reported 17 minutes ago

  Grok has a connector for Gmail and I use it for search. Also to find and organize invoices, things I may be overpaying for, ways to cut costs, etc. It’s infinitely better than the broken built in Gmail search

• 
  Labrador_Dad (@DM_Se7en) reported 22 minutes ago

  @gmail Just really make the app better and for crying out load. Fix the search!

• 
  Biju Peter (@bijupeter) reported 26 minutes ago

  You reset your phone and try to login to #gmail or #microsoft app, they send security message to the same app. How the hell are we supposed to access this? Same with #linkedin as well. What kind of security system is this?

• 
  B. Warnecke (@BenisBoy14) reported 26 minutes ago

  @ITContractorsUS @nvidia Dude. Fix the grammar issues before making a screenshot. GMail even marks it all for you.

• 
  Mingke (@Mingke) reported 30 minutes ago

  @ylecun @nicochristie People don’t choose iOS or proprietary models because of their closedness. They are ordinary people who cannot — or simply do not want to — spend time managing compute (iOS) or approximations of intelligence (models). They want it done by someone else, and are therefore willing to pay. Revenue is the outcome of a service: someone provides value by helping you get things done. In the case of iOS, that service is managing compute resources. Users pay for that and if they believe they can manage better themselves and are willing to take on the effort to maintain it, they can always switch to Android because of its exsitence. In the end, a more a philosophical stance: how should human society be organized, centralized or individualistic? Governance exist to hold people together (so we don’t end up killing each other just because goals differ and communication is inefficiently low-dimensional) and to enable further exploration of a unified goal. But managed by whom? iOS believes they can manage better than the mass population. Android doesn’t even argue on that axis. It simply claims you can do what you want — freedom as the form factor. SMTP is open, yet most people still choose Gmail. They choose to have the service done by Google rather than setting up their own email server using open protocols. Using revenue to measure open source versus closed source is therefore meaningless. The better metric is whether freedom is safeguarded as a genuine choice — not whether freedom is actually exercised by everyone. That's the value of open source, as an option keeping the entire system permissionless. Yet promoting open source is another form of elite protection for the masses: the majority need not exercise the choice, and may not even fully understand its trade-offs. The value lies in the fact that the door remains open.

• 
  Cortex (@0xCortexl) reported 31 minutes ago

  ANTHROPIC RELEASED AN OPEN STANDARD THAT IS DOING TO AI TOOLS WHAT HTTP DID TO THE WEB - AND 90% OF DEVELOPERS STILL DON'T KNOW IT EXISTS every AI integration today is custom code - and when the model changes everything breaks MCP is the first protocol built for how models actually think - the model asks what can you do, the server responds, the model picks what it needs with zero glue code Google Calendar, Notion, Gmail - old way is 3 integrations of 500 lines each - MCP is 3 servers and the model figures out the rest the same way HTTP let any browser talk to any website - MCP lets any model talk to any tool save and learn this today - developers who understand MCP now will build the infrastructure everyone else runs on