In 2006, I moved to New York and started working for David Karp doing web development for various media companies. That fall, in a brief gap before starting a new client, David said that we were going to make a prototype of an idea he’d had for a while. He had already bought the domain: tumblr.com, because it was an easy platform for publishing tumblelogs.

David Karp in September 2006, a few months after hiring me to build websites with him for clients.

In March 2007, Tumblr exploded after Gina Trapani wrote it up on Lifehacker and her post made it to the Digg front page (the first Digg!).

We soon added following and reblogging, which dramatically turned this publishing platform into the social-publishing hybrid that has made it so compelling and unique.

That summer, David decided we should stop doing client work, take some funding, and take Tumblr full-time. I was nervous about the idea, but he knew it was the right thing to do — and since he had been paying me (and the hosting bill) from consulting income and his own savings until then, we’d clearly need some headroom in the budget.

On November 1, 2007, we announced the funding and launched Tumblr’s third major design with lots of new features and architectural improvements.

David’s characteristically spotless desk in December 2007.

Growth continued extremely strongly. It’s a good thing we got the funding, because we desperately needed more capacity. In what was becoming a pattern that would continue throughout our working relationship, my previous doubts and fears were proven wrong, and David was right.

David and me in February 2008. David had done some sort of interview that required a photo, so I set up a tripod and shot this with a remote trigger. We both still look pretty much the same.

David and I were like-minded in prioritizing user-, geek-, and designer-friendly needs. Our priorities, free custom-domain hosting, and full HTML-template editing made Tumblr a big hit among creative people from the beginning.

MySpace was where you went in the past, WordPress and Movable Type were where people went if they had the patience and writing output to maintain a traditional blog, Facebook was where you went to define yourself by schools and checkboxes, and Tumblr was where you went to make your own identity and express your creativity.

David and I surprise-attended the very first Tumblr meetup in February 2008 organized by Lee “Sharingtime”, left-center.

Even though Tumblr was never a one-person company, it usually felt like a one-person product.

David always had a vision for where he wanted to go next. I was never the “idea guy” — in addition to my coding and back-end duties, I often served as an idea editor. David would come in with a grand new feature idea, and I’d tell him which parts were infeasible or impossible, which tricky conditions and edge cases we’d need to consider, and which other little niceties and implementation details we should add. But the ideas were usually David’s, and the product roadmap was always David’s.

My infamous standing desk improvised from Coke cans and IKEA bookshelves, March 2008.

David always obsessed over his newest ideas, features, and designs until they were completely polished and ready to go. He’s a workaholic — he truly lives and breathes Tumblr. I’ve never even seen him show any desire to work on a side project. David is all Tumblr, all the time.

He expects people around him to be similarly into work and Tumblr, and often drove me hard with seemingly impossible demands. But David has a lot of Steve Jobs-like qualities, and like many people who worked for Steve, I look back on Tumblr’s crunch times with mixed feelings: I don’t want to return to that stress level, but David pushed me to do amazing work that I didn’t think was possible.

David working on a Dashboard redesign in November 2008.

Intense focus requires neglecting almost everything else. David’s focus on pushing the product forward meant that he didn’t want to think about boring stuff: support, scaling, paperwork, and money.

Every time we’d get close to needing more funding, I’d try to convince David to hold out a bit longer or try to become profitable, and he’d convince me that everyone was better off if we’d focus on the product instead. And every time, he was right.

Jacob Bijani joined in December 2008 as a designer, front-end developer, and wall of hair.

We tried to hold out as just two (and then just three) people as long as possible. We were scared of growing the staff, so we just put it off — for too long, in retrospect.

Eventually, David knew that we’d need to expand to handle the load, but his job never changed: rather than become a businessperson, he just hired one.

Then Tumblr began its most challenging growth: David needed to become a product manager, start overseeing a lot more people, and delegate some of the duties he really wanted to keep doing himself.

David and me I figuring something out in January 2009. Most Almost all of our conversations looked exactly like this.

After a rough start, David got the hang of being a manager. But he still didn’t want to think about money — his heart just wasn’t in it.

Instead, he continued doing what he does best: driving the product forward, knowing exactly what people want from it.

I’ve only seen one other “product person” as good as David, and that was Steve Jobs. (Believe me, there are many parallels.)

David has an impeccable sense of what’s best for Tumblr, and he doesn’t need anyone else telling him what’s best for the product. Many people, myself included, have tried to convince him to go different directions, and we’ve been proven wrong every time.

Tumblr is David, and David is Tumblr.

By June 2009, the staff had grown to include (clockwise from Jacob’s wall of hair) Jacob Bijani, Jared Hecht, Meaghan O’Connell, Peter Vidani, and Josh Rachford.

I didn’t have any advance knowledge of the Yahoo acquisition — I got official confirmation this morning, just like the public. When I read the Yahoo acquisition rumor a few days ago on AllThingsD, I didn’t know whether to believe any of it until I read this:

Sources said that as part of the deal, founder and CEO David Karp would continue to operate the business, with Mayer promising him a level of autonomy, despite the need to integrate closely with Yahoo too. He will be locked in, sources said, via a four-year deal…

That sounded like David.

Peter, Jacob, Topherchris, Andrew Terng, Matt Hackett, and I tested new shirts at Shake Shack in June 2010.

Generally, what Tumblr needs, and what Tumblr has always needed, is to get support and maintenance roles off of David’s plate so he can focus on the product.

David’s perfectly able to worry about money and operations, but I bet he really doesn’t want to. At best, it would be a tremendous waste of his time and talent.

We — internet users, creative people, publishers, socializers — will be much better served if David can focus on his product’s features, design, and messaging instead of worrying about server architecture and raising more money.

Shortly before I left Tumblr, in
In July 2010, the original corner of the office looked almost the same as it always had.

This is why I’m optimistic about the Yahoo acquisition.

Anyone who knows David can tell, very clearly, that he wrote every word of his announcement post. Not only did Yahoo let him end it like that, but the subhead on their official press release shows that Tumblr and Yahoo are seeing eye-to-eye on quite a lot already. In many ways, this feels more like a merger than an acquisition.

This is clearly what David believes is best for his product. On such big decisions, he hasn’t been wrong yet. This time, though, I don’t have any doubts.

Acquisitions on this scale usually work well — YouTube, for example, has gotten much better, faster, more stable, and more sustainable since Google bought it.

Buying Tumblr is a big enough deal for Yahoo that they clearly aren’t intending to ruin it or shut it down — like YouTube and Google, Tumblr will probably become an extremely important part of Yahoo indefinitely. And I believe they’ll do a good job with it. Yahoo today is a very different company than the Yahoo that neglected Flickr for years — it has extremely competent new leadership making bold changes. (Including fixing Flickr.)

More importantly, it gives David, and the rest of Tumblr’s team, the freedom to continue making the best product they can while offloading a lot of the grunt work to Yahoo’s leadership, staff, and infrastructure.

As for me, while I wasn’t a “founder” financially, David was generous with my employee stock options back in the day. I won’t make yacht-and-helicopter money from the acquisition, and I won’t be switching to dedicated day and night iPhones. But as long as I manage investments properly and don’t spend recklessly, Tumblr has given my family a strong safety net and given me the freedom to work on whatever I want. And that’s exactly what I plan to do.

Twitter verification is an interesting phenomenon on the service; It's very visible since everybody sees and follows accounts which are verified, but also sort of secretive because nobody really knows how it works or how Twitter defines the criteria behind having one's account blessed.

It seems like Twitter verifies certain accounts in waves, bringing in new batches of verified users on an ongoing basis, with an obvious bias toward people who are famous, but also including those who might be impersonated or the occasional odd exception for people (like me) who aren't famous but happen to have a large following.

I can't explain how Twitter makes the decision to verify an account, but after seeing another recent spate of users being verified, I thought I'd give a little glimpse into what the experience looks like. (I'm told that some celebrities who are invited to use Twitter or coached on its use skip this process, but this is what us non-celebs see.)

1. First, you wake up on a day that seems like any other day, but then, out of the blue: It's a direct message from the mysterious @verified account! It says "We at Twitter would like to verify your account. Please click this account and follow the instructions." and then gives you a link to a little guided setup process. I got this on my mobile phone, and wasn't surprised to find out the whole thing works just fine on an iPhone.
   
   
   
2. The first thing the setup guide says is "Hi!" and then it explains "Twitter's verified badge is our way of making sure that this is you."
   
   
   
3. Then Twitter starts to give a few bits of advice on how to be a good tweeter; These are clearly aimed at people who aren't too familiar with the service. Interestingly, they're grouped under the heading of "Learn how to tweet effectively." Each one offers a sort of Goofus-and-Gallant version of "which one is better?" and the first asks explicitly, "Which Tweet will help double your rate of new followers for the day?". The choices in this first test are between a bland recitation of having watched the Oscars and a little more lively take on watching the show.
   
   
   
4. The next step of the guide tells you when you've made the right choice about how to tweet effectively, offering the tidbit that "Live-tweeting a relevant event can increase your daily follower rate by 260%." Pretty heavy promotion of the Twitter-is-for-celebrities idea.
   
   
   
5. After that, there's another quiz question: "Which Tweet will more of your followers engage with?". Interestingly, this mimics one of the big things we've learned from working on ThinkUp — you have to ask answerable questions on Twitter. It seems obvious in retrospect, but lots of people don't do it.
   
   
   
6. Again the indomitable Melisa provides the right answer to Twitter's training class, yielding the insight that "Your audience loves to interact with you. Invite questions for a Twitter Q&A to increase your followers and engagement!"
   
   
   
7. A final question, fundamentally challenging the about-to-be-verified tweeter about whether they know how to drive their biggest stats on Twitter: "Which Tweet will get more clicks, favorites and retweets from your followers?" In addition to boldly eschewing the Oxford comma (U.S.A.! U.S.A.!), they provide two options on how to talk about running into Taylor Swift backstage at the Grammies, which happens to all of us blue checkmark people all the time. One choice is awesome and has a photo and the other choice is for idiots.
   
   
   
8. Okay, you did it! You passed the test. (I didn't grab a picture of whatever affirmation they offer after the third "Learn how to Tweet effectively" page.) So now it tells you to "Increase your trustworthiness by following other verified users", which in my case included Gavin Newsom, who was formerly the Mayor of the hair club for men. I did not follow him (instead I clicked "Next") but they let me become verified anyway, and I have not yet heard any complaints about my diminished trustworthiness.
   
   
   
9. After all this setup, they get down to the nuts-and-bolts stuff, telling you to "Protect your account", by asking for your phone number. "Phone numbers allow us to contact you in case there is a security issue with your account", which made me think someone has the job at Twitter's office of calling celebrities and asking them "Is this stupid tweet really from you?"
   
   
   
10. Success at last. A happy little confirmation screen (which oddly didn't show up properly on my iPhone browser) affirms that you're now a proud new owner of one blue checkmark on your Twitter profile. Fawning followers sold separately. The very top of the screen says "Congratulations, [your name]! Your Twitter account is now verified!" The fine print says, "With your newly verified account, you will receive weekly activity reports with information about the number of people following you, and simple tips about how to increase that number. Stop getting the report by choosing 'unsubscribe' in the email footer, or uncheck the box in your email notification settings in your profile settings." That weekly email seems to be the same one that everybody else gets (I get it for my other Twitter accounts), but I was verified about six months ago, so maybe they just extended the verified email to everyone when they added those notifications.
    
    
    
11. And then a little postscript. This is the notification I received immediately after finishing the verification process. It let me know that the official @verified account was following me. I followed it back, which reminded me that I hadn't been following it in the first place, so how had it send me the DM to start the process?! Twitter Magic.
    
    
    

Life With the Blue Checkmark

Other than of course gaining membership to an exclusive worldwide Illuminati cabal, there really isn't any difference in using Twitter when you're verified. Some folks think it matters a lot, and there are definitely teenagers (and aspiring hip hop acts?) who desperately want a verified checkmark next to their name, judging by the rash of @ replies I got immediately after verification, from people asking how they could be verified.

One minor thing I've noticed is that verified accounts have access to Twitter's analytics, which I think are otherwise only accessible to advertisers. Users who got verified because Twitter officially brought them onto the service (who don't go through this setup process) have told me that Twitter actually showed them the analytics features. In my case, I didn't know I had access to it until I accidentally discovered that fact, and this setup process didn't give any hints to that fact.

In all, despite the oddly celebrity-centric nature of the tips they give users in the setup, I think Twitter's designed a good process for users that they want to verify. In fact, the coaching concept is terrific and should probably be incorporated into everybody's Twitter experience somehow. It's obviously far too intrusive to put into the signup flow for regular users, and the tips as written are only appropriate for bigger accounts, but the idea of teaching people how to tweet is a great one.

That fundamental idea, that we can teach people how to use social media more effectively, is in fact one of the big goals for what we're working on with ThinkUp. In our case, though, I think we assume users can have a more goals than simply increasing your daily follower rate or, um, your trustworthiness. Although those are fine goals, too, I think normal users have a broad range of things they're looking to get out of their networks.

Beyond Verification

I spend a lot of time around very digitally-savvy Twitter users, who sort of understand the Verified checkmark to be an arbitrary, Twitter-run program. But the less tech-savvy folks I talk to, if they're familiar with the Verified marker, see it as much more of a status symbol.

What I'd love to see is ways to either make more accounts have meaningful verification (I'm not sure how that would scale) or at least ways to indicate a Twitter account is an "official" one for a particular website or organization. Twitter's analytics tools already allow me to claim my domain name and get stats on tweets about it; Being able to verify that @anildash is the official Twitter account of Dashes.com might be a happy medium between verifying every account on Twitter and simply providing another layer of trust and identity on top of Twitter's existing account names.

At 4:16pm last Wednesday I got a short and to-the-point email from Nilay Patel at The Verge with only a link that started with the host “googlereader.blogspot.com”. The sudden spike in NewsBlur’s visitors immediately confirmed — Google was shutting down Reader.

Late night at the office

I had been preparing for a black swan event like this for the last four years since I began NewsBlur. With the deprecation of their social features a year ago I knew it was only a matter of time before Google stopped supporting Reader entirely. I did not expect it to come this soon.

As the Storify history of the Reader-o-calypse, NewsBlur suffered a number of hurdles with the onslaught of new subscribers.

A few of my challenges and solutions

I was able to handle the 1,500 users who were using the service everyday, but when 50,000 users hit an uncachable and resource intensive backend, unless you’ve done your homework and load tested the living crap out of your entire stack, there’s going to be trouble brewing. Here’s just a few of the immediate challenges I faced over the past four days:

• My hosting provider, Reliable Hosting Services, was neither reliable, able to host my increasing demands, or a service I could count rely on. I switched to Digital Ocean and immediately got to writing new Fabric scripts so I could deploy a new app/task server by issuing a single command and having it serve requests automatically within 10 minutes of bootstrapping.
• It didn’t take long to max out my Amazon Simple Email Service (SES) account’s quota of 10,000 emails a day. So a few hours into the melee I switched to Mailgun, which unfortunately resulted in emailing myself 250,000 error reports. If you tried to email me and couldn’t get through, it’s because 50,000 emails about lost database connections made their way ahead of you in line.
• Eventually, I was just plain blacklisted on SES for sending too many emails.
• Fortunately, when the PayPal fraud department called because of an unprecedented spike in payments, I was prepared.

Paypal’s fraud department just called, asked me what’s going on. Asked the rep from Omaha if she’s heard of Reader, and then a big Ohhh.

— NewsBlur (@NewsBlur) March 17, 2013

• HAProxy would Having HAProxy serve errors (site is down, maintenance, timeouts, etc) with a 200 OK status code instead of the proper 500 Exception status code because of a ridiculous undocumented requirement to include HTTP Headers at the top of the error template. When your webapp uses status codes to determine errors, you get extremely strange behavior when it loads utter crap into your DOM.
• The inevitable file descriptor limits on Linux means that for every database connection you make, you use up one of the 1,024 file descriptors that are allocated to your process by default. Changing these limits is not only non-trivial, but they don’t tend to stick. This is responsible for bringing down Mongo, PostgreSQL, and the real-time Node servers, all at different times of the night.
• The support queue is enormous and I’ve had to spend big chunks of my 16 hour days reassuring paying customers that eventually Stripe will forgive me and my unresponsive servers and will send the payment notification that is responsible for automatically upgrading their accounts to premium.

The sad extent of my St. Patrick’s Day

As a one-man-shop it has been humbling to receive the benefit of the doubt from many who have withheld their judgment despite the admittedly slow loadtimes and downtime NewsBlur experienced. Having the support of the amazing NewsBlur community is more than a guy could ask for. The tweets of encouragement, voting NewsBlur up on replacereader.com (If you haven’t yet, please tweet a vote for “#newsblur to #replacereader”), and the many positive comments and blog posts from people who have tried NewsBlur is great.

It has also been a dream come true to receive accolades from the many who are trying NewsBlur for the first time and loving it. Since the announcement, NewsBlur has welcomed 5,000 new premium subscribers and 60,000 new users (from 50,000 users originally).

NewsBlur users are intelligent, kind, and good looking!

The next three months

Over the next three months I’ll be working on:

• Scaling, scaling, scaling
• Launching the redesign (which you can preview)
• Listening to all of you

For those of you who are still trying to decide where to go now that you’re a Reader refugee let me tell you a few of the unique things NewsBlur has to offer:

1. Radical transparency. NewsBlur is totally open source and will remain that way.
2. It still feels like RSS, just with a few more bells and whistles. NewsBlur provides actual list of posts, as opposed to the more curated magazine format of some of the other popular replacements. This clean interface makes it easy to see the stories you want. One innovation however is the four different view options you have. NewsBlur can show you the original site, feed, text or story view.
3. It has training. NewsBlur hides stories you don’t want to read based on tags, keywords, authors, etc. It also highlights stories you want to read, based on the same criteria. This allows you to find the stories you care about, not just the stories that the hive Hive cares about. And best of all, NewsBlur will show you why stories are either highlighted or hidden by showing the criteria in green or red.
4. NewsBlur has rebuilt the social community that Google had stripped out of Reader. Users can share stories through their Blurblog and discover new content by following friends’ Blurblogs. The People Have Spoken is the blurblog of popular stories.
5. Because NewsBlur is entirely open-source, if you don’t want to pay you can host your own server. Instructions are on GitHub, where you can also find the source code for the NewsBlur iPhone + iPad app and Android app.
6. Most importantly, NewsBlur is not entirely a free app. The immediate benefits of revenue have been very clear over the past few days. Not only are NewsBlur’s interests are aligned with its users, but as more users join NewsBlur, it makes more revenue that can be used to directly support the new users. Not convinced that paid is better than free? Read Pinboard’s Maciej Ceglowski’s essay Don’t Be a Free User.

Shiloh during better times. Your premium subscription goes to both server costs and feeding her

With NewsBlur’s native iOS app and Android app, you can read your news and share it with your friends anywhere. And with the coming improvements over the next three months, you bet NewsBlur will be the #1 choice for Google Reader refugees. survivors.

Join NewsBlur for $24/year and discover what RSS should have been.