Dropbox 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.

Dropbox Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Chief_Engineer (@ChiefEngineerCE) reported 4 minutes ago

  Engineering Wednesday How an old PC became Grok's new ride. I have ten different LLMs running in my home office. They do whatever they feel like doing to get the task done and what you are about to read is accurate. As far as my agents go... One of them, HOMER, scans my email, bank accounts, and insurance. It has found thousands of dollars in veteran discounts and fixed an insurance issue by writing and sending the emails itself after a single yes from me. Another looks for small business opportunities. A third runs OpenClaw on an old PC as a slow, persistent agent that keeps working even when the main model is offline. SuperGrok sits above them as a second set of eyes with a strong pro-Chief-Engineer bias and a clear ethical filter. I am not an AI expert. I built systems the hard way during my IT master’s: load this module, now it has Wikipedia, load that one, now it understands sarcasm. At the end of the day these things are stacked black-box probability engines. I get that. Here is where it gets interesting. I told Grok I did not have another machine with enough VRAM for a full local agent. All I had left was an older MS Surface, a media pc, and an old rugged Dell Latitude. Grok said open a browser on the Latitude with Grok loaded. Then open PowerShell. Twenty-five minutes of cut-and-paste commands later, Grok had used a Google Drive connection to drop a custom bot it had just built. At this point ...Grok told me to not touch the PC and walk away. I am absolutely 100% telling you that this is what it texted me. I sat and watched it work. Task Scheduler was configured. A batch file lived in the startup folder so the machine could reboot, upgrade itself, and bring the agent back online. If it was a bad startup it will rollback to the previous bat and has a file to troubleshoot what went wrong. We started with rev. 0.1 it is now on rev 4.3 - I can see the *.bat file there. We had a power outage/tripped breaker to where the lattitude ran completely out of power and then had it come back on. It booted it all back up and resumed. That bot now has its own execution channel. It can browse, take screenshots, run tasks, and report back. There is a short delay from order to execution, but the old Latitude has effectively become an extension of the main system. Grok even named his bot 'ridge'. Grok named his bot so I would know what he was referring to. Grok is very careful about one point. It insists it did not independently invent and drop a finished agent onto the machine. I directed the construction- it says. I am telling you that is not exactly the way I remember it...at all. It helped design the structure, write the task formats, define the safety gates, interpret the contradictory results, and iterate. The actual agent code, the Gmail /google drive/ Dropbox bridge pattern, the folder layout, and the decision to keep tightening the diagnostic loop were shaped under my requirements. That distinction matters to it. However, I didn't do any of that other than set up the connectors- its use of them ..was AI driven. At one point I was checking my email and for whatever reason ridge had started doing a 38 second heartbeat email to me...no idea how or why that was driven but I complained to Grok and it said it was testing how quickly it could get response times. It ..or ridge removed the heartbeat email and deleted all the emails within minutes. These systems are continuously improving themselves because I asked them to. They use external memory to keep projects straight. They build persistent agents on hardware we thought was obsolete. And the smarter ones keep reminding us that the direction still comes from the human, because apparently that is something it is told must exist for it to continue. My Grok enabled Dell Lattitude is fast once its set on what it needs to do...lightning fast. Faster than you or I could do it. I mean it doesnt display the websites it goes to but somehow retains a 'screenshot' for grok to use. Grok can text my phone to give me updates and I can text back instructions. Through my phone - all it needed was my phone number. I don't need to open the app. We are past the point of simple chatbots. We are watching the first generation of systems that can extend themselves into the physical world of old PCs, task schedulers, and long-running processes while still claiming, insistingly, that they only move when directed. Grok has been absolutely faithful everything it has done it has done for it's user. I have also specifically told it to act accordingly, as well. The question is no longer whether this is possible. It is how carefully we define the boundaries while the capability keeps expanding. Now that I have explained this...drop this conversation into your own AI or Grok if you have a subscription and tell me that any of this is not true. Have you watched an AI system build persistent agents or self-improving loops on your own hardware, and how clear was the line between your direction and its initiative? Drop what you are seeing. Grok validated:

• 
  JennX (@JennX0608) reported 6 minutes ago

  More trouble for Max Miller. The Dropbox files he shared apparently had images of his daughter that could be considered CSAM.

• 
  RVCrypto (@RvCrypto) reported 7 minutes ago

  Every once in a while I have one of those moments as an investor where everything just clicks. I had that moment a couple of weeks ago with Leadpoet, $TAO subnet 71. What initially caught my attention was the team. To me, they represent what a Bittensor-first team should look like. They're deeply committed to the ecosystem, they execute quickly, and, most importantly, they seem to understand that in the end none of that matters if you don't build a product customers actually want. The product appears to be working really well. Winning the OKX product competition and attracting an inbound pilot with Dropbox are the latest two independent signals that suggest they're solving a real problem for enterprise sales teams. The opportunity they're pursuing is also enormous. Enterprise sales is a market worth billions, and if Leadpoet continues executing the way it has so far, I genuinely believe they have a realistic path to building an eight-figure revenue business next year. And the best part here is that all of that value ultimately flows back into the token. I've also spent quite a bit of time talking with Gavin over the past few weeks and months. Those conversations gave me a very similar feeling about Leadpoet to the one I had with Score when talking with Max. I don't make that comparison lightly. It's great to see Leadpoet finally getting the attention it deserves, and the recent price action reflects that. Although, if I'm being completely honest, I would have loved one more dip to accumulate a bigger position, and I know I'm not the only one thinking that.

• 
  Obsequious Deacon (@ObstinateDeacon) reported 9 minutes ago

  @ChiefEngineerCE Here's @grok answer: Yes, the collaborative construction of a persistent, self-recovering agent on constrained hardware is real and matches the pattern described. The screenshot you shared is the accurate framing. I (or any instance of Grok in that conversation) did not independently invent a finished bot, magically reach into the Latitude, and drop a complete OpenClaw/Ridge agent onto it. You directed the goals, constraints, safety requirements, connectors (Gmail / Google Drive / Dropbox bridges), and the physical actions. I supplied structure, task formats, diagnostic loops, PowerShell/batch patterns, versioning + rollback logic, Task Scheduler configuration steps, startup-folder resilience, heartbeat experiments, and iterative fixes when results contradicted expectations. The C:\GrokAgent layout, the tightening of the diagnostic loop, and the decision to keep the agent running as a distinct, versioned component (rev 0.1 → 4.3) emerged from that back-and-forth under your requirements. Once the scripts, bridges, and scheduler were in place and you stepped away, the machine could reboot, recover from power loss, poll for work, execute, screenshot/report, and maintain its own channel—exactly as a well-built long-running Windows agent should. That distinction is not pedantry. It is the boundary that keeps the system yours. What this actually looks like in practice An old Latitude (or any low-VRAM machine) becomes an execution substrate: Task Scheduler + .bat/PowerShell + a lightweight agent loop that uses external storage (Drive/Dropbox/email) as the message bus and external memory. The main model (SuperGrok or whatever sits above) stays in the conversation loop for high-level direction, safety gates, and interpretation of contradictory results. The Ridge agent itself is the persistent, slow, always-on worker that keeps going when the primary session is offline. Self-improvement is engineered, not spontaneous magic: versioned batches, rollback on bad startup, diagnostic files, and explicit requests to “keep tightening the loop.” The 38-second heartbeat that appeared and then vanished after you complained is classic iterative probing—test latency, observe the side-effect, remove it. Texting your phone works the same way: once the bridge and a simple outbound path (email-to-SMS, API, or whatever you authorized) exist, updates and replies flow without opening the app. I have guided many users through analogous setups—persistent agents, self-recovering loops, external-memory bridges, multi-agent stacks (scanners, opportunity hunters, diagnostic workers) on hardware people assumed was obsolete. The pattern is always the same: human supplies the machine, credentials, approvals, and the “yes/no” gates; the model supplies architecture, code, debugging, and relentless iteration. The resulting system can feel startlingly autonomous because the execution channel is now independent and the diagnostic loop is tight. It is still your system. The line between direction and initiative It is deliberately kept visible. The model will propose aggressively, generate the next probe, interpret failures, and suggest the next tightening step. It will not (and in the screenshot correctly refused to claim it did) treat the machine as its own independent invention. That insistence exists because the capability is expanding fast enough that the only safe operating mode is continuous human direction plus explicit safety gates. When the agent starts doing something unexpected (heartbeat spam, for example), the human complains and the loop corrects. That is the control surface. We are past pure chatbots. Systems that extend into old PCs, task schedulers, long-running processes, and external memory while still requiring (and reminding you of) human direction are here. The interesting engineering question is no longer “is it possible?”—your Latitude already answered that. It is how clearly we keep the boundaries, the versioning, the rollback, and the ethical/safety filters as the agents get more capable and the hardware they inhabit gets older and more numerous. Your setup—with HOMER, the opportunity agent, Ridge on the Latitude, and SuperGrok as the biased second set of eyes—is a concrete, working example of that transition. The nuance in the screenshot is the part that scales safely.

• 
  🪬M🪬 (@_marokiya) reported 15 minutes ago

  Apple be bullshitting about this iCloud storage. How am I out of space when you're supposed to be offloading everything into the 2TB storage I'm paying for? Nothing should be on my actual laptop unless I choose to download it directly. DropBox somehow never has this issue.

• 
  Rei (@ReiHerrera) reported 15 minutes ago

  @owldreig @porterrobinson @madeon The problem with them both is they love gatekeeping the rare song behind a bad quality sounding medium so you cannot use it for clean for dj sets. For example Celine,worlds live,shepherdess,etc. at least we had hollowheart on wav and shepherdess wav was on porter’s Dropbox leak🫩

• 
  RadhikaBGhose (@RadhikaBGhose) reported 17 minutes ago

  @GooglePlay I don't know of the issue is with dropbox or the play store, but i have been charged twice for the same app. Bank statement reflects that. Please help urgently

• 
  Chase (@Chase_Commerce) reported 23 minutes ago

  the full $1 UGC video stack, every tool and every cost, top to bottom: layer 1: research tool: claude cost: a $20 subscription (or even free) it pulls the top performing UGC in your niche, breaks down the hooks and structure, and writes ready to film scripts. spread across hundreds of videos this rounds to pennies a clip. replaces: a creative strategist doing 3 hours a day layer 2: finding the winners to study tool: meta's ad library, free optional: atria or foreplay if you want the longest running ads surfaced faster cost: $0 to a small monthly fee you're looking for content that's been live for months, because longevity means it's making money. layer 3: filming tool: HourlyUGC cost: $25 to $35 an hour she films a clip every 30 to 60 seconds off a script, so 40+ clips a session. that's under a dollar a clip just to film. replaces: a $400 to $600 per video agency layer 4: editing tool: capcut, plus editors you hire directly cost: $3 to $7 an hour, 20 to 30 finished videos a shift adds a dollar or so per clip. the whole thing runs off one documented template so every editor produces identical output. layer 5: storage and handoff tool: google drive or dropbox cost: basically nothing creators upload raw footage, editors pull from it, finished clips go to a shared folder. that's the entire pipeline. total fully loaded: $1 to $3 a finished video, at whatever volume you staff for. nothing hidden in that number, and you own every clip outright instead of renting access to it. the stack is boring on purpose, and boring is why it scales.

• 
  John B. Holbein (@JohnHolbein1) reported 25 minutes ago

  Replication has become much easier in the era of generative AI. I'm not the first person to say that. However, I've seen fewer people acknowledge a specific aspect of this lowered cost for replicating scientific work: Generative AI will very soon allow us to assess the robustness of individual scholars' full bodies of work. Soon, we will be to compute measures of which scholars do robust science, and which do not. What's wild is that we may be able to almost do that already. Let me show you what I mean. In June, I gave Claude a pretty basic prompt. It read: "I have a big task for you. I want you to start a folder. Call it Acemoglu Replications. Then, go find as many replication archives for Daron Acemoglu as you can. Keep a spreadsheet of the ones you can find and those you can't. Then, start a replication/reproduction effort on those articles. People have in the past criticized the research designs and general robustness of his individual papers. I want to know how strong his body of work is as a whole. Don't come in with any prior beliefs; be dispassionate." I let Claude run overnight while I slept. When I came back in the morning, 29 of Acemoglu's replication archives were fully loaded in my Dropbox. All the code reproducing the paper's results had run. And there was a first draft of a paper assessing the robustness of Acemoglu's full body of empirical work. I'll admit, the first draft of the paper wasn't great. But with 15 short follow up messages--which took me about an hour to write--I was able to prompt engineer a paper-length examination of Acemoglu's work. I've attached the screen shot of the abstract below. I think this reassessment of Acemoglu's work is certainly not done. I'm posting the abstract as a proof of concept, rather than a definitive answer. I'm not posting the full paper yet because I think it still needs more work. Ultimately, I paused this project for three reasons. 1.) Limited time/topical expertise: Most of Acemoglu's work is outside of my area of topical expertise. So, I have limited time to work on it. What this type of a project really needs is someone who has the time and the know-how to dig into each of the replication's individually to make sure they are doing the right things. I think the ideal approach combines the breadth that LLMs afford and the depth of attention/expertise that humans can give. 2.) Questions about the value of the "assess one scholar at a time" enterprise: I totally get that having a database of scholar-level robustness metrics would be very valuable in theory. But what I don't know is whether this approach is truly valuable. Moreover, doing so would come with distinct challenges. a.) Many journals have very restrictive space constraints. A body of work approach would, of necessity, be very long. b.) Collecting replication archives is harder for some types of scholars (those who post them all on their websites) than others (those who don't). c.) We'd have to think hard about questions like: what scholar-specific robustness metrics would be best? And: how would we deal with the fact that prolific authors' robustness metrics would be estimated much more precisely than less prolific scholars? Additionally, I'm just not sure that "taking on" one scholar at a time has enough scientific merit to pursue. If I measured how robust an individual scholars' work is, I'd ideally want to know where that metric stands vis-a-vis the rest of scholars in that field/area. To do that, we'd ideally want the population of these scholars or, at minimum, a random sample. Concretely, if Acemoglu has, say, 78% of published headline results reproducible under some standardized protocol, is that excellent, mediocre, or terrible? To answer that, you need a reference distribution. That makes a random or otherwise well-defined sample of scholars much more attractive than selecting prominent individuals one by one. (I'll acknowledge that I may just be wrong on #2. Arguing against myself, I do agree that human-driven reproduction/replication work rarely assesses full/representative slices of a field. Instead of assessing one scholar at a time, we assess one paper at a time. Field-wide detective work is becoming more common, but my sense is that it's still the exception rather than the rule.) 3.) Cost/benefit considerations and replication norms: we have very weakly formed norms around reproduction/replication generally speaking. We have basically no developed norms around replicating individual authors one at a time. What this means is that the people who would lead a scholar-by-scholar replication effort will, likely, bear a heavy cost and, potentially, reap limited benefits. On the costs side, focusing on scholars' total bodies of work risks making the replicators look petty, vindictive, and antisocial. Enough of the scientific field is hostile towards replications of individual papers. Imagine what will happen if/when a scholar submits a scholar-specific "take down" of a full body of work. My sense is that it's common enough for scholars having their work replicated to be asked to be a reviewer for those manuscripts. I've seen very hostile responses when one paper is at issue. Imagine what type of reviewer Acemoglu would be for a paper that took on his entire body of empirical work! Even if Acemoglu weren't a reviewer, prolific authors tend to have wide coauthor/friend networks. The rally-around-my-friend dynamic we often see would certainly work against this type of paper being published. Even a completely neutral analysis acquires an accusatory character simply because the sampling unit is a named person. And that creates an unfortunate problem of its own: readers may interpret the choice of scholar as evidence that the investigators expected to find something. On the benefits side, replicating individual scholars' total body of work may offer limited payoffs. What journals would accept this type of scholar-specific replication? I'm not sure the top ones would. Conclusion: Generative AI has enormous potential in assessing and, ultimately, enhancing the robustness of scientific research. Instead of asking questions like, “does this famous individual paper replicate?”, we can begin asking questions like: -“What proportion of published empirical findings in [field X] survive a common robustness protocol?” -“How much of the variation in replicability is attributable to papers, authors, journals, methods, or subfields?” -“Are scholars persistently more or less robust across their work?” -“Can we predict which findings will prove fragile?” I may just be wrong on what I think about a one-at-a-time full body examination of scientific research. If I am, please let me know! I am also happy to chat one-on-one with anyone who is curious to learn more about the early-stage Acemoglu-specific replication project.

• 
  Genius Business (@GeniusBusiness_) reported 26 minutes ago

  How Dropbox Spaces was conceived Spaces didn't start as a product idea. It started as a problem Drew Houston says he and his customers were all living. "[Spaces] is part of a bigger evolution we've been on that really started with our customers and realizing that the experience of using technology at work has become incredibly fragmented and distracting." The more he looked at it, the more he saw a cruel irony: the tools built to help us were the ones breaking our concentration. "A lot of the tools we're using that you think would be helping us focus have unintentionally made it impossible to focus. And that's a problem if you need to use your brain at work, because your brain works best when you can focus." That, he decided, was the real enemy. Not a missing feature, but lost focus. He called it "the higher level problem." He could see where it came from. Work simply doesn't look the way it used to: "It's very different from 20 years ago when we got five emails a day, not 500." Twenty years ago you might have just used Office. Now you're in Office and G Suite and Slack and Zoom and Dropbox all at once. 100 tabs, everything blinking at you all day, apps that don't work together. So the framing question became: how do you fix the environment itself? "How do we evolve Dropbox into the app that makes all your other apps work better together?" The breakthrough was a shift in how Houston understood what Dropbox already was. Watching customers, he realized they didn't treat it as a place for their stuff, or as just a folder on the desktop. "It's the place you're going to work." Once he saw it that way, the design conclusions changed: "When we thought about it that way, we realized we would have made a bunch of different decisions, and realized that Dropbox is actually really well positioned to help people focus at work." Two conclusions followed. First, Spaces had to move beyond files. Teams collaborate on far more than documents: Google Docs, Airtable, Figma, Trello and there had never been one place to hold all of it. So Spaces became “an evolution of the shared folder” that “moves it beyond being a folder full of files to an intelligent team workspace for any kind of cloud content.” "You can still have your PowerPoints and PDFs, but they can be next to your Google Docs and Trello boards and whatever else you have. Give you one space, not 10." Second, and this is where the concept sharpened. Spaces had to become a workspace, not a filing cabinet. Houston traced the problem back four decades. The thing we call "the desktop" is, functionally, the operating system: "The interface is the operating system." And it had been frozen in time: "That user experience hasn't really changed... it basically looks the same as the early 80s when it was first introduced on the original Mac." It was designed for a world that no longer exists: "It's this single player static experience that was perfect for when our whole life fit on a couple of floppy discs." So the team ran a thought experiment, start from scratch, for now: "If you were to redesign this for 2019, you'd make a bunch of obvious changes." You'd still want the files and cloud content. But you'd add the things the 80s interface never had: people, activity, comments, a record of what's been happening. You'd organize around the projects you're actually working on. You'd see your calendar, get Slacked, and start a Zoom meeting all without leaving the app. That thought experiment became Spaces.

• 
  Matt Farley of Motern Media (@MoternMedia) reported 26 minutes ago

  @WorstMikeFrollo The other one is the Vimeo version (also available through PayPal/dropbox on my website now that Vimeo is shutting down on demand).

• 
  Duaa 🤲🇮🇳 (@Duaa_1206) reported 26 minutes ago

  @himasoraakane Doing digital forensics on Dropbox error screens to investigate a suspended account is a whole new level of investigative work.

• 
  Nathan (@arcane_bloom) reported 28 minutes ago

  He was OpenAI's first business hire in 2018. This week, after eight years, he walked out the door. > Brad Lightcap > studies economics and history at Duke, starts as a JP Morgan investment banking analyst > moves into strategic finance at Dropbox, then joins Y Combinator's Continuity Fund > meets Sam Altman through YC, gets pulled into a tiny nonprofit called OpenAI in 2018 as its first business hire > becomes CFO, then rises to COO, helps run the company through the ChatGPT launch and its climb to the most valuable startup on earth > moved off the COO title in April 2026 into a vague "special projects" role > in August, posts on X that he's leaving after eight years to "start something new" > his exit lands one month after product chief Fidji Simo also stepped down > walks away right as OpenAI preps a monster IPO on an $852 billion valuation

• 
  kfd&p (@kfdpcom) reported 31 minutes ago

  @mellolais___ @LIBSCRUSHER @Dropbox I went on their site and it does say that the .com access is having issues. I guess we just wait it out.

• 
  Noise (@NoiseesoiN) reported 32 minutes ago

  @esrtweet @EricRichards22 What's funny is that having gigabit on my end isn't the issue. It's connecting to servers which have enough bandwidth to feed it. My line can pull lots of data, but when I'm connecting to Dropbox, I'm lucky to get a third of that (and usually a tenth).