Dropbox status: access issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Dropbox reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Dropbox users through our website.

• Errors (60%)
• Sign in (20%)
• Website Down (20%)

Live Outage Map

The most recent Dropbox outage reports came from the following cities:

City	Problem Type	Report Time
Nottingham	Errors	22 days ago
Guayaquil	Website Down	22 days ago
Flumet	Errors	1 month ago
Irapuato	Errors	1 month ago
Bournemouth	Sign in	3 months ago
Paramaribo	Errors	4 months ago

Full 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:

• 
  Mapacho (@Mapacho111) reported 5 minutes ago

  @MajorianBTC But that’s the problem. Bitcoin in its current state is a decentralized Dropbox. Bitcoin failed and there’s no alternative.

• 
  OneToothTeXan (@OneToothTeXan) reported 6 minutes ago

  I'm so sorry I left my zipper down and my sanity got loose. If found: Men, there's a dropbox. Women: please return to original source.

• 
  Mikemira (@storiesbyohama) reported 6 minutes ago

  Just imagine getting accepted into the most exclusive startup club on earth… Then being told you have two weeks to find a complete stranger to as your partner. That’s exactly what happened to Drew Houston in 2007. He had the idea for Dropbox. He had a rough demo. @ycombinator liked it. But @paulg was clear... Single founders rarely make it. You need a co-founder. Right now. Drew’s friends couldn’t join. Time was running out. What would you do? He put out the word. A mutual friend connected him to a quiet MIT student named Arash Ferdowsi. They had never met. They sat down in the student center. Talked for about two hours. About code. About the problem. About the future. At the end of that conversation Arash said yes. He dropped out of MIT the next week with only one semester left. Two weeks later they walked into the YC interview together. They got in. The rest is history: a company that became worth billions. It looked reckless. It felt like a shotgun wedding. Yet it worked because both were all-in from the first conversation. I’ve studied hundreds of startups that never made it past the idea stage. Most founders wait too long for the “perfect” partner. They overthink chemistry. They protect their equity. They miss the window. You can’t wait for certainty. Sometimes the right co-founder is the person willing to jump with you before the proof exists. The speed of that decision can be the difference between staying a solo dreamer and building something real. What would you risk in two weeks if the right person walked in?

• 
  Ishita Sethi (@Ishita__Sethi) reported 9 minutes ago

  @LoganOpSec that dropbox bit is the part that sticks. recovery without revoking access is just half a fix.

• 
  Ryan Babbs (@buckybabbs) reported 12 minutes ago

  Hello @Dropbox, Just this week I have decided to switch cloud service from Google Drive to you and less than 1 week I'm having an issue. I run a wedding film business and am trying to move the same 200GB file to two separate editors for two separate kind of films and one of them is encountering a "link temporarily disabled" message. I tried to research online and made a new folder and copied over the original folders contents and cannot share that either as it says I've exceeded bandwidth limits. This is maddening as I thought I'd picked the good plan (3TB) and was all set. Please advise asap as I need these files in my editors hands stat.

• 
  Abhishek Singh (@0xlelouch_) reported 15 minutes ago

  System design question. How would you design Dropbox-style file sync with conflict handling? Constraints that make it interesting: 1) Multi-device edits while offline, then reconnect hours later 2) Large files (GBs), but common case is small diffs; do you chunk + hash + resumable upload? 3) At-least-once events from clients; duplicates and retries are normal 4) Need per-file causality: version vectors? server-assigned sequence? something else? 5) Conflicts: rename vs edit, delete vs edit, concurrent edits on same bytes; what gets auto-merged vs forked as conflicted copy? 6) Metadata vs content planes: directory tree ops must be atomic-ish, blobs can lag 7) End-to-end integrity: how do you detect corruption and avoid re-upload storms? 8) Fast convergence: target <5s to see changes on another device, but mobile battery + bandwidth are limited

• 
  davidsong (@bitplane) reported 15 minutes ago

  Has @Dropbox been breached again? I just got this but it can't be me. At least I hope not. I have a randomly generated, unique password that's made in an offline password manager, I used it once on this machine when I first set it up. If I'm getting hacked the rest of you are in serious trouble.

• 
  Craylor (@craylor) reported 19 minutes ago

  Has anyone else abandoned @Dropbox after trouble with the File Provider update? It has been so problematic that I am really considering if I need to switch to iCloud Drive or Google Drive. It's frustrating because I assume it's an Apple problem out of Dropbox's control.

• 
  Juan Pabro (@ironwoodreserve) reported 21 minutes ago

  @BitPaine @Dropbox Is now the worse time to try to login? I was a sent 2FN right as I was getting ready to click on what to use to login. It seemed way too fast for 2FN.

• 
  More Gravy (@lotsmoregravy) reported 21 minutes ago

  @FFT1776 Deputize our military and send them to literally every last damn polling station and dropbox in the United States of America. Every last damn one. Give them the authority to handle **** on the spot. Problem solved.

• 
  Arthur Terrell Vaughn (@ArthurV36405381) reported 23 minutes ago

  @patmendoza6745 I know not of any of those specifications, but I still can’t figure out what the issue is with persistent memory, why can’t I just have a dropbox online that it can constantly access every single chat and conversation?

• 
  Rei (@ReiHerrera) reported 24 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🫩

• 
  Fraser (@iamfra5er) reported 24 minutes ago

  THIS GUY WANTED A PLACE TO STORE HIS PASSPORT WITHOUT DROPBOX READING IT so he built an encrypted vault app for himself in a weekend and it's now doing $5k/mo zero startup cost. 85% margins. no ads. just SEO written by an AI agent trained on his emails the agent finds trending topics on reddit every single day, writes an article, translates it, posts it google indexes it in days. 500-600 daily visitors. 4% convert to app store downloads. all running on free cloudflare then ASO does the rest — he translated the app into 36 languages and ranks #1 for "duress vault" in the US app store 80 downloads a day. 9% conversion to paid. completely autonomous most founders obsess over their first 10 customers but this guy got banned from every reddit community and said whatever, I'll just let the robot handle distribution he's an ex-google security engineer who raised hundreds of millions for his last startup so he knows what terrible UX looks like in security apps every competitor either has bulletproof security with unusable UI or easy UI with trash security he just combined both and called it done doesn't even spend time on this app. works on 4 projects at once. lets coding agents build while he plans the MVP is identical to the final product because he built exactly what he wanted for himself no pivot. no customer discovery calls. just "I need this, maybe 10 other people do too" now he's testing tiktok and youtube not even for this app but just to learn distribution for the next one

• 
  GHOST 🌙 (@ghosstty_) reported 26 minutes ago

  OPUS 5 + HIGGSFIELD CAN TURN A 6-QUESTION FORM INTO A $35K WEBSITE. IN ONE SESSION. FOR $4 IN TOKENS. no mockup. no wireframe. no mood board. six answers from the client. that's the input. a live website is the output. here's the form. here's what each question does. and here's why agencies charge $35K for what this produces in one session. → QUESTION 1: "WHO IS THIS SITE FOR?" not "describe your target audience in 500 words." one sentence. "CFOs at mid-size SaaS companies looking to switch billing providers." this one answer sets the tone, the copy angle, the visual weight, and the CTA hierarchy. an agency runs a two-hour discovery call to get this. I get it in a google form on monday. → QUESTION 2: "WHAT SHOULD A VISITOR DO?" book a demo. buy the product. join the waitlist. one action. this kills scope creep before it starts. no "maybe we should also add a blog and a careers page." one page. one goal. one conversion. → QUESTION 3: "SHARE 3 SITES YOU LIKE AND SAY WHY." not "what's your brand aesthetic?" nobody can answer that. "I like stripe because it's clean. I like linear because of the motion. I like notion because it feels simple." three links. three reasons. that's the design system seed. → QUESTION 4: "WHAT MAKES YOU DIFFERENT FROM COMPETITORS?" one paragraph. sometimes one sentence. this becomes the headline. the subhead. the entire above-the-fold story. a copywriter would interview the founder for an hour. this question does it in 30 seconds. → QUESTION 5: "SEND YOUR LOGO, BRAND COLOURS, AND ANY EXISTING ASSETS." a dropbox link. a google drive folder. sometimes just three hex codes and a png. this grounds the design in reality. no inventing a brand from scratch. no "let's explore some directions." → QUESTION 6: "WHEN DO YOU NEED IT LIVE?" not a timeline negotiation. a date. "next friday." done. that's when it ships. → WHAT HAPPENS NEXT friday night. six answers go into the pipeline. Opus 5 takes the answers and builds. design system. responsive layout. copy. CMS. all from the spec those six answers created. Higgsfield generates the hero media. product shots. clips. matched to the brand. saturday I review. adjust. polish. sunday morning - walkthrough link in the client's inbox. → WHY THIS WORKS because $35K was never the cost of building a website. it was the cost of figuring out what to build. discovery calls. alignment meetings. three directions. two rejected. scope changes. revision rounds. all of that is just a slow, expensive way of answering six questions. I ask the questions upfront. the client answers in 10 minutes. the machine builds from the answers. $4 in tokens. one session. same site. the full system - the form, the pipeline, the stack, and the pricing model - is in the article below. reply "FORM" and follow me - I'll send you the full playbook.

• 
  Indragie Karunaratne (@indragie) reported 29 minutes ago

  I used the first Dropbox beta back in 2006 and was sold right away - I’m still a paying customer of the product today, 20 years later. But there hasn’t been much innovation in this space since then and we’re in the midst of a broad shift in how people interact with computers. I’m excited to back this great team and see what a modern take on this problem looks like!

• 
  David Carcelli (@DavidCarcelli) reported 29 minutes ago

  @Dropbox dude if you guys don’t get of the Dave is requirement I’m done. I make music and I also use a cpap. I have no problem finding something better than this nonsense.

• 
  Ricky Shah (@RickyShahatty) reported 30 minutes ago

  @Claudio_PNW @tax_birdie @AMandoSch Miller wanted the Dropbox released publicly. It is entirely up to the attorney to screen it. A bad client should make an attorney double down their efforts.

• 
  Olivia Jane. 🦇 🗡️ 🩸 (@livtheripper) reported 32 minutes ago

  @HazeOfBlue84 Haha, I get it! Thankfully someone here at the office set it up and all I had to do was re-sync Dropbox! I had no issues at all, so I'm thankful.

• 
  Nathan (@arcane_bloom) reported 37 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 37 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.

• 
  DAX : TV (@daxtv) reported 40 minutes ago

  @0x00_dev @ProtonMail have been using dropbox for years- never an issue - thought it might be smart to move to Proton Drive - but not now me thinks - hey ho :)

• 
  Just Matthew (@MatthewP279348) reported 43 minutes ago

  @Rani_Rant_Fest @iGardon Doesn't change the fact that they aren't showing the prompt, so the result is valueless. I don't read people's google shares. Put it in dropbox if you want me to read it. Lastly, calmatters is a CA bureaucracy, of course they will lay down cover for their corrupt gov.

• 
  Mohamed irfan (@heyIrfan) reported 48 minutes ago

  The Marketing Strategy That Actually Works Most people make the same mistake when building a product: They try to sell before they prove that they can solve a real problem. Think about companies like Google and Amazon. They didn't start by saying "Give us your money, and we'll make you rich." They solved problems people already had. That's the foundation of good marketing Don't start with selling. Start with solving. 1. Solve a real problem When you're building something, your product will always look amazing to you. Your idea feels perfect because you built it. But that doesn't mean the market wants it. The only way to find out is to Talk to real users. Understand their problems. Find out what they're currently doing. See whether your product actually makes their life easier. Don't assume your idea is valuable. Let the users prove it. 2. Give value before asking for money Don't immediately push your product. Give people something useful. Your solution should help them Save time. Save money. Reduce effort. Solve a painful problem. If you genuinely create value, selling becomes much easier. You're no longer saying "Please buy my product." You're saying "This solves a problem you already have." That's a completely different conversation. 3. Don't compete only on features Your competitor has 10 features. You build 15. Then they build 20. And now you're stuck in an endless feature race. Instead, compete on value. Ask: "How much better can I solve the user's problem?" The differentiation shouldn't just be "We have more features." It should be: "We create more value for the customer." 4. Let people try before they buy Give users a way to experience your product. Especially with AI products, you don't necessarily need to give everything away for free. Give enough access for them to understand the value, while keeping usage manageable. Then collect feedback. But don't blindly follow every piece of feedback. If someone says: "Change the button color." That doesn't necessarily mean your product needs to change. Look for feedback about the actual problem and experience. 5. Don't forget the people who already showed interest Someone visited your website. Someone signed up. Someone tried your product. Someone talked to you. Those people are valuable. Don't immediately try to sell to them. Talk to them. Understand why they came. Understand what they liked. Understand what stopped them. And if they leave, ask why. Because the person who leaves may know something you don't. They might reveal the hidden problem that helps you improve the product. 6. Price based on value Don't blindly make your product extremely expensive. And don't make it extremely cheap either. Your price should be: Affordable for the customer + sustainable for your business. Being cheaper than competitors can help, but price alone shouldn't be your strategy. If your product saves a company $1,000 every month, paying you $100 can feel like a great deal. That's because the customer isn't really buying software. They're buying the value your software creates. Look at Google Drive Google Drive is a simple example of value-first thinking. The problem: We need to store files. We could keep everything on a pen drive. But then we have to: Carry the device. Manage files manually. Worry about losing it. Move files between devices. Share files manually. Google Drive makes this much easier. Your files are stored online. You can access them from different devices. You can share a link. You can control whether someone can view, comment, or edit. And you don't have to build your own storage system. There are competitors too: Dropbox, iCloud, OneDrive, and others. So Google Drive isn't valuable simply because "it stores files." It's valuable because it solves the bigger problem around storing, accessing, managing, and sharing files. And Google gives users a free amount of storage so they can experience the product. You can try it. You can upload files. You can share them. You can experience the features. Then eventually you may reach the storage limit and think: "This is actually useful. I don't want to delete my files. I'll pay for more storage." That's the important part. They didn't need to convince you with a sales pitch. They let you experience the value. And once you experience real value, paying becomes an easy decision. The strategy is simple: Find a real problem → Solve it → Give value → Let users experience it → Talk to users → Improve the product → Then monetize. Don't sell first. Create value first. Because when you solve a real problem, the product starts selling itself.

• 
  🔩⚾️ (@john_dough) reported 48 minutes ago

  @irisivanyi @nickcamsmith If someone has a very fancy and complex portfolio site, they are probably not working very much. My portfolio is 10+ years out of date because I never have the time to update it. I just send a Dropbox folder of images of new work to people these days. ¯\_(ツ)_/¯

• 
  Hago Community (@HAGOCommunity) reported 49 minutes ago

  AI Internal Search Agent: An Intelligent Agent for Searching Company Information Many companies struggle with information being scattered across multiple systems and files. Policies may be stored in Google Drive, documents in SharePoint, conversations in Slack or Microsoft Teams, customer data in a CRM, while internal procedures may be stored in Notion or Confluence. When an employee needs specific information, they may have to search in several places, ask a colleague, contact a manager, or open multiple files before finding the correct answer. This is where an AI Internal Search Agent can help. This agent is an AI-powered system that can search across different company data sources, understand an employee’s question, and provide a direct answer based on the internal information available to that employee. How Does the Agent Work? The agent can be connected to the systems and platforms used by the company, such as: Google Drive SharePoint Notion Confluence Slack Microsoft Teams CRM systems Internal databases PDF files Internal documents Company policies Standard operating procedures Employees can then ask questions in natural language instead of manually searching through multiple systems. For example: “What is the company’s travel expense reimbursement policy?” Or: “Where can I find the latest version of this customer’s contract?” Or: “What are the steps for adding a new customer to the system?” Or: “Who is responsible for this account, and what was the latest update?” The agent searches the sources the employee is authorized to access and provides the most relevant answer. The Problem It Solves The main problem is usually not that the company lacks information. The problem is that employees do not always know where that information is located. An employee may spend time: Searching through multiple folders. Opening several documents. Reading old conversations. Asking coworkers where information is stored. Trying to identify the latest version of a document. Searching across different business systems. This creates unnecessary delays and wastes employee time. Instead, the employee can simply ask the AI agent and receive an answer within seconds. A Practical Example Imagine an employee wants to know the process for purchasing new software for their department. In a traditional workflow, the employee may search through emails, ask their manager, and browse company folders until they find the correct policy. With the AI agent, the employee could simply ask: “What is the process for purchasing software that costs more than $5,000?” The agent could search the company’s internal policies and respond: “Purchases above $5,000 require approval from the department manager first. The request must then be submitted to Procurement and approved by the Finance department.” The agent can also provide a link or reference to the original policy document used to generate the answer. Searching Customer Information The agent can also be used to search customer-related data. For example, a sales employee could ask: “What was the latest agreement with customer ABC?” The agent could search the CRM, internal notes, documents, and customer-related conversations before providing a summary. For example: “The latest meeting with the customer was on August 12. The customer is interested in the Enterprise plan and requested a revised proposal before the end of the month.” This allows the employee to understand the current status of the account without manually searching through a long history of notes. Searching HR Policies Employees can also use the agent to get answers about internal HR policies. For example: “How many annual vacation days do employees receive?” “What is the remote work policy?” “How do I request time off?” “What is the process for business travel?” Instead of sending these questions repeatedly to the HR department, employees can receive answers directly from the AI agent based on official company policies. Supporting New Employees One of the most useful applications of an AI Internal Search Agent is employee onboarding. New employees often have many questions, such as: “How do I request a laptop?” “How do I access the internal system?” “Where are the team files located?” “Who approves expenses?” “How do I submit an IT support request?” The AI agent can act as an internal assistant throughout the onboarding process and provide immediate answers to these questions. Respecting Employee Access Permissions One of the most important features of the agent is permission management. Not every employee should have access to every piece of company information. For example, some documents may contain sensitive information related to payroll, contracts, human resources, finance, or executive management. The agent should therefore respect each employee’s existing access permissions. If an employee does not have permission to access a specific document, the AI agent should not use that document when generating an answer. This allows the company to provide intelligent internal search while maintaining appropriate data access controls. Showing the Source of the Answer The agent should not only provide an answer. It should also show the source of the information whenever possible. For example: “According to the company travel policy updated on May 3…” The employee can then open the original document and verify the information. This helps reduce the risk of employees relying on outdated or incorrect information. Detecting Outdated or Conflicting Information The agent can also be designed to identify conflicting information. For example, it may find two different documents containing different instructions about the same company policy. Instead of selecting one version randomly, the agent could alert the employee or administrator: “There are two documents containing different instructions regarding the remote work policy. The most recent document was updated in June.” This can also help companies improve the quality of their internal knowledge management. Moving From Search to Action The system can be developed to do more than simply search and answer questions. For example, an employee may ask: “How do I add a new customer?” The agent can first explain the required steps. The employee can then say: “Start the process.” The agent could create a checklist, create a new record in the CRM, send a request for the required documents, and notify the employee about the remaining steps. At this point, the system moves from being an AI Search Agent to becoming an AI Operations Agent. Example Inside a Sales Team A sales representative could ask: “What are the most important things I should know about this customer before the meeting?” The agent could search the CRM, previous notes, proposals, and communications before creating a summary that includes: Company size. Products the customer is interested in. Date of the latest meeting. Previous objections. Estimated deal value. Recommended next steps. This allows the sales representative to prepare for the meeting without spending significant time searching for information. Example Inside Customer Support A customer support employee could ask: “How was this problem solved in the past?” The agent could search previous support tickets and the company knowledge base to find similar cases and show the solutions that were previously used. This can reduce ticket resolution time and help new support employees solve customer problems more efficiently. Data Sources the Agent Can Connect To The agent can potentially connect to many different systems, including: Google Drive Microsoft SharePoint Slack Microsoft Teams Notion Confluence Salesforce HubSpot Dropbox OneDrive ERP Systems CRM Systems Internal Databases PDF Documents Excel Files Company Policies Employee Handbooks Customer Records The more organized and up-to-date the company’s information is, the more useful and reliable the agent becomes. Benefits for the Company An AI Internal Search Agent can help a company: Reduce the amount of time employees spend searching for information. Reduce repetitive questions between employees. Make policies and procedures easier to access. Help new employees become productive faster. Improve knowledge sharing across departments. Reduce dependence on individual employees who know where everything is stored. Make customer and project information easier to access. Reduce the time required to find the correct documents. Improve the speed of decision-making. How Can the Company Measure the Agent’s Value? The company can measure the value of the system by calculating how much time employees previously spent searching for information. For example, if 200 employees each spend 20 minutes per day searching for files or asking coworkers for information, that represents more than 66 hours of employee time every day across the company. If the agent can significantly reduce that time, the financial value of the system becomes easier to measure. The company can also track: Average time required to find information. Number of questions answered by the agent. Number of manual searches avoided. Reduction in repetitive questions sent to HR and IT. Customer support resolution time. Time required to prepare employees for meetings. What the Final Agent Workflow Could Look Like The employee asks a question ↓ The agent understands the request ↓ It identifies the relevant data sources ↓ It checks the employee’s access permissions ↓ It searches the company’s internal data ↓ It selects the most relevant information ↓ It provides a concise answer ↓ It shows the source of the information ↓ If authorized, it can also perform the requested action In this way, an AI Internal Search Agent becomes a private intelligent search engine for the company, allowing employees to access internal knowledge quickly instead of wasting time searching across files, systems, and conversations.

• 
  Brandon (@BJohnsonxAR) reported 53 minutes ago

  @DropboxSupport I’m having issues logging in. When I try to login it’s making me sign up when I already have a paid account. This is happening on both the app and my browser. @Dropbox

• 
  Alvin (@Alvin1492840) reported 55 minutes ago

  Kill the startup apps that have been draining your battery since day one. She opened System Settings → General → Login Items & Extensions. 14 apps were set to launch automatically every time he turned on his Mac. Spotify. Zoom. Adobe Creative Cloud. Google Drive. Microsoft Teams. OneDrive. Dropbox. A VPN he used once. A screenshot tool he forgot about. A calendar widget. And 4 more he didn't recognize. Every one of them was running in the background 24/7 consuming RAM, CPU cycles, and battery life whether he was using them or not. She said: "You turn on your Mac and within 30 seconds, 14 apps are fighting for resources before you've even opened your first document. Your fan spins up because your CPU is processing a traffic jam of apps you're not using. Your battery dies by 2pm because half your power is going to background processes you don't see." She removed 11 of the 14. Kept only the ones he actually needed at startup. The Mac booted in half the time. The fan stayed quiet. The battery lasted 3 extra hours. She said: "Check this list right now. If you see apps you don't use daily, remove them. They've been silently eating your Mac alive since the day you installed them."

• 
  Saurabh | Celsius 233 (@Celsius233Books) reported 56 minutes ago

  @colemickens @Dropbox lenovo and security issues...takes one back to the 2015 Superfish scandal where lenovo was visually scanning every single webpage you visited to sell ads

• 
  Mansi 👩‍💻 (@MansiCodez) reported 58 minutes ago

  Solution of yesterday’s question: Design Google Photos: the part after the boxes “Hash it and put it in S3” fails the interview. Two phones compress the same sunset differently. Same photo. Two hashes. Two rows. You just built a worse Dropbox. The system needs three IDs, not one. client_upload_id — generated on the device before the first byte moves content_hash — hash of the exact bytes you received asset_id — the thing the user sees in the library Uploads are sessions. Library entries are assets. Blobs are renditions. If you collapse those into one key, retries, edits, and shared albums all collide. 1. Retries must be idempotent on the client, not on the filename Phone goes offline mid-flight with 612 shots, 40 already half-uploaded. Each photo gets a client_upload_id the moment it enters the queue. Chunks are uploaded against that ID. Commit is PUT /uploads/{id}/complete. Same ID + same bytes → same session. Server returns the existing asset. Late packet after commit is a no-op. Filename + timestamp is not an ID. Camera roll and AirDrop will mint two. 2. Exact dupes are content-addressed. Near-dupes are reconciled. After commit: look up sha256(bytes) if it already exists for that user (or the shared album’s owner set), attach the new upload to the existing asset_id do not create a second photo The 28 shared “Goa 2026” shots that are almost-but-not-quite the library copies will miss on sha256. That is expected. Run a cheap perceptual hash (pHash / dHash) + capture time + camera model from EXIF. If distance is tiny and captured within a few seconds, mark as near_duplicate_of and do not show two tiles. Keep both blobs if you must; hide one in the UI. Two devices, two compressions, one photo in the grid. 3. The library is a set of assets + tombstones. Not last-write-wins. Delete in Delhi must beat a pending upload in Mumbai. Every mutation carries: asset_id op: upsert | delete | restore actor_id (device or user) logical_ts (per-actor Lamport or hybrid logical clock) A deleted asset gets a tombstone that outlives the pending queue. When the flight-mode phone finally flushes those 40 half-uploads, the server sees: upload commit for an asset that already has a newer delete → commit the blob if you want, do not resurrect the tile. Refresh in Mumbai cannot show a photo Delhi just deleted, because the change feed is “tombstone wins over delayed create,” not “whoever wrote last.” 4. Shared albums are references, not copies Partner adds 28 photos to Goa 2026. The album stores {asset_id, added_by, added_ts} — not a second blob, not a second library row. Adds and removes are a small CRDT: add(asset, actor, ts) remove(asset, actor, ts) Two devices adding the same asset = one membership row. Phone sync finishing a second later cannot wipe the partner’s 28 photos, because there is no “replace the whole album document.” Last-write-wins on the album JSON is how photos vanish. 5. An edit is a new rendition, not a new photo and not an overwrite User crops + filters while the original is still processing. Rules: original blob is immutable edit creates rendition_id with parent_asset_id library still shows one asset “current view” pointer moves to the latest rendition history is a list of renditions / edit ops, not 12 full-resolution copies by default If you overwrite the original, face clustering and search lose their source. If you mint a new asset, the user now has original + edit as two photos. Both are wrong. Storage stays sane because you store: original (once) derived thumbs / display sizes lazily, keyed by asset_id + transform not every intermediate crop as a first-class photo 6. Upload path and ML path must not share a lock “Beach sunset with Priya” in minutes, not overnight, also not on the upload critical path. Commit path only: durable bytes asset row appear in library + album enqueue jobs Workers (thumbs, embeddings, face cluster, labels) are async. Search index is eventually consistent. The UI can show the photo immediately with “processing” on faces. If clustering blocks upload, you built a spinner, not Photos. Face identity hangs off asset_id, so an edit does not orphan Priya. The new rendition inherits the parent’s cluster and gets re-checked, not reset. 7. Sync is a checkpoint + change feed, not “download the library” Each device stores last_applied_ts. Server gives a stream: new assets, new renditions, album membership, tombstones. That is how 62,000 existing photos plus 612 offline shots plus 28 shared adds converge without a full rescan, and why a deleted photo does not climb out of another device’s queue. The one-line design Client-generated upload IDs stop retries from cloning. Content hashes stop exact clones. Perceptual reconcile stops “same sunset, different JPEG.” Tombstones stop resurrection. Album CRDTs stop last-write-wins from deleting the partner’s night. Edits are renditions under one asset. ML is a consumer of commit, never part of it. Boxes for S3, CDN, Kafka, Redis are table stakes. This is the part that decides whether you designed Google Photos or a photo-shaped file dump.

• 
  Sridhar Katakam (@srikat) reported 59 minutes ago

  Tips to new @OmarchyLinux users from a new Omarchy user 1. If you are coming from a Mac and have installed Omarchy on a Windows machine: Your keyboard most likely has start button and then the Alt button. Swap these so Super key becomes command where the left thumb usually rests. I asked Claude Code to do this for me. 2. If your mouse has a scrollwheel: Super + mousewheel scroll up and down switches desktop spaces. This is besides the usual Super + <number> and Super + Tab / Super + Shift + Tab to switch spaces. In addition to these, I have AI set up Ctrl + Alt + arrow. 3. Missing Alfred/Raycast? Install omacast. 4. If your laptop is connected to an external monitor with bluetooth mouse and keyboard and you find that the computer is not waking up (after you walk away for a while) when you press any key or move the mouse, ask AI to fix this for you. 5. The default font size of text in Claude Code is 12px which is small for my aging eyes. Ask AI to change it to 16px. 6. Todoist does not have a native app for Linux and so if you miss its Quick Add feature, ask AI to set this up. It added the functionality and assigned Super + Q to it. Later, I came to know there's a community plugin for it. But I am happy with what Claude did. 7. If you use Dropbox, chances are, the total available space in your account is not accurate because it is hardcoded in Omarchy at this time. Ask AI to fix this for you. 8. If/when the fan in your computer is spinning fast and is noisy, ask AI to find out the reason and fix. In my case, it turned out to be Performance profile in the Lenovo laptop. Claude changed it to Balanced and now it's all fine. 9. Get familiar with keyboard shortcuts.