Welcoming Flightmode beta into the world

Today we released Flightmode. That means that you can now download our baby in the iTunes app store :)


We’re calling it a soft launch, which means we’re not gonna make a splash about it. We’re still very much in feedback mode as we’re still not at the quality level that we want to be at. Quality is incredibly important to us.

This is an MVP — a minimum viable product — the smallest version of the app that we think you’d be happy with. We started this work 2 years ago. Along the way we’ve realised that the minimum bar in today’s tech world is high. People expect quality and are generally suspicious towards new apps.

So we spent 2 years working on various prototypes, getting feedback in a lot of different ways without spending too much money. The end result from all those iterations and feedback is that we’ve built 3 separate systems:

  1. Flightmode, the app
  2. Alexandria, the database
  3. Sia, the algorithm

We want to tell you about these three systems. More importantly we want to tell you about the journey that got us to this point and why we’ve built what we have.

Flightmode, the app

It all started with what’s now called Flightmode. The core idea was always an app that lets you easily read quality journalism. It’s just that we underestimated the importance of the word easy.

Our very first prototype was called Edify. Edify means “instruct or improve (someone) morally or intellectually.” Back then we thought it was a very clever name. There was a snootiness and snobbishness to Edify that we’ve since gotten rid of.

With 5 tabs, Edify was complex. You could follow journalists and publishers, you could discover articles by topics, you could follow each other’s recommendations, you could share articles with specific groups, and you could discover articles using a Tinder-like function. Edify was information overload personified.

Edify morphed into Vouch. Vouch focused on the social aspect of Edify because we realised that people rely on other people for their content recommendations. Vouch was gonna be this new cool social media for smart content. We wanted to introduce gamification where people would collect points called Impacts based on how many people were impacted by their recommendations.

Vouch also tried to deal with information overload by only giving 6 pieces of content at the time. The intention was to be small, but the prototype remained incredibly complex.

Vouch became Impactable when we realised that we couldn’t use Vouch because of associations and naming issues. (Yes, we actually thought Impactable was a good name for a while.) Impactable was even more focused on the importance of the content and the impact it would have. And it was still way too complex.

We then made a critical shift from focusing on the importance of the content to focusing on relaxed enjoyment. At that point we called ourselves Hammock. Hammock was still a social media, but it was packaged differently and took a marked step towards simplicity.

One key feedback in the life of Hammock was that people didn’t want the hassle of having to follow people. So we added a Tasting Menu, which was content recommended by the top recommenders on Hammock.

That tasting menu was the first step in dropping the social angle that we had kept for so long. Eventually we dropped the social angle entirely and instead focused solely on information overload. We essentially made the Tasting Menu the centre of our product.

One key reason we were slow to drop the social angle is that we knew how pervasive the filter bubbles were. We realised this long before the Trump election brought the filter bubble conversation into the mainstream. We had noticed these bubbles from all the social prototypes we created all the way in the beginning of this project.

Our main learning about filter bubbles is that they are not as one-dimensional as you might think. It’s not as simple as red vs. blue. The truth is that you and I are in different bubbles; we simply like to read about different stuff. So in creating the Tasting Menu ourselves, we knew that we’d need to cover the full spectrum of content to provide good recommendations for even a small group of close-knit people.

We therefore decided to first test our ability to provide recommendations manually using 3rd party apps. Once we realised that we could meet that challenge consistently we started doing more proper work on the visual design. We focused 100% of the visuals on restoring attention spans and dealing with information overload. And Hammock was too chunky to pronounce so we renamed it Flightmode.

Looking back, Edify was essentially just an all-encompassing, complex RSS feed. Vouch, Impactable and Hammock were just people sending each other URLs that we re-packaged in a specific format. In all these apps, readers would have had to do much more work than they were actually willing to do.

The challenge with Flightmode was that it shifted the burden onto us. We needed to build a much more sophisticated system that encompassed our understanding of your preferences for journalism. Dealing with information overload in the age of filter bubbles is the essential challenge of the systems behind Flightmode. So we started off by creating Alexandria.

Alexandria, the database

Named after the great library in Egypt, Alexandria follows roughly 15,000 journalists, thinkers, specialists and intellectuals for article recommendations.

These people are organised into +70 channels. Each of those channels represent a filter bubble that we discovered during all those prototypes. Suddenly all that time manually stalking people’s Twitter and Facebook profiles during the Vouch days came in handy.

Alexandria organises all those recommended articles by word count (this is vital in today’s attention-zapped media environment) and by quality (this is the secret sauce…).

Alexandria pulls some 500 quality articles every day. Unfortunately the best articles are far too long, but journalists and publishers will one day realise that they need to compress those great articles into fewer words. That’s where demand is and supply will obviously follow.

Every time we work in Alexandria we are mesmerised. It’s like a smart Google full of great content, all organised by authors, publishers, topics, channels, and length. Future investors will probably consider it the real jewel of our company. Alexandria’s potential is huge.

Despite its greatness, Alexandria on its own is kind of useless. Alexandria is pure information overload. So we created Sia:

Sia, the algorithm

Named after one of the Egyptian gods associated with intellect and writing, Sia decides what goes from Alexandria into your personal Flightmode. Sia gets an understanding of your interests and attention span. It asks you about some broad categories and then figures out the details on its own.

Many people are worried when they realise that we’ve built an algorithm for journalism. They’ve been boxed in by algorithms before, getting the same Kardashian story over and over again after a single moment of weakness. Or they’ve been lead into a filter bubble that made them unaware of the whole picture.

Sia will be different because it has a different intention. Most algorithms want to engage your most unconscious attention. Sia wants to find the edge of your comfort zone and then push that zone.

We believe Sia is working well when it gives you 2 articles on your biggest interests and 1 article at the very edge of your interests. That 1 article that you didn’t even know you’d find interesting, but somehow opened you up to something new.

Sia aims to find that golden balance between the expected and the unexpected. We believe this balance is critical in order to keep your interest and create good media habits.

An option would be to give you 3 random articles across the full spectrum, but all our feedback and observations suggest that this would be a terrible product for most people. Our odds of keeping your interest would simply be too low. Alternatively we could give you 20 random articles, but then we’re back into Edify territory.

To keep Sia in line with those intentions, we’ve added a strong human element to the whole process.

The human editor & conductor

As a human balance to Alexandria and Sia, there’s me, Anders, named after my Norwegian ancestors. I will work double shifts as editor and conductor.

I’m qualified to do this because I’ve spent A LOT of time reading journalism. Not just these last 2 years, but also the 10 years before it. I believe that I’ve read close to the full spectrum of journalism.

As editor, I will look through every article that comes out of Alexandria. Or at least I’ll do that until I’m replaced by an artificial intelligence or a person that does it better than me. I make a decision whether to boost, approve or delete every article that appears in Flightmode. Here’s what I do:

I delete the descriptive news updates and biased event promotions that might have made it through the system. This is rare.

I boost the articles that have an unexpected angle. I boost articles that I believe you might find interesting even if you are not familiar with that niche. I boost those articles that might be your bridge into an unfamiliar bubble. I boost articles that cover topics that aren’t trending. I boost articles that I believe are important for the public debate. I boost articles that I would really love to read. I boost articles that make my heart jump.

And then I approve the rest. Odds are you’ll only receive those articles if you’re into that specific channel. At least that’s the intention.

My other role is the conductor. As a conductor I play with the variables in the algorithm. I try to balance the instruments; keep the trumpet and the violin in balance.

Part of that work will be personalised for you, at least in the beginning of this journey. So if you send me a request, I can link you up with a channel you’re curious about. Or alternatively, block a channel that you simply don’t want to read about. That’s totally fine, enjoyment is our priority.

As a conductor I’ve also gone into Alexandria and manually given a bump to 100+ journalists that I know deliver quality content. That list will grow with time as we get an even better view of who the top writers are. Great writers tend to deliver great content consistently.

And I’ve given a bump to 200 publishers that I also want to promote. These include:

  • traditional quality names like the Atlantic, the Economist, Foreign Affairs, and Harper’s.
  • bold critical voices like openDemocracy, The Intercept, ThinkProgress, and Zero Hedge.
  • traditional hobby magazines like National Geographic, Scientific American, IndieWire, and Sports Illustrated.
  • consistent providers of lesser-known quality journalism like Glamour, Marie Claire, Buzzfeed, Teen Vogue, Playboy, GQ, and Vice.
  • interesting international voices like Haaretz, The Wire, Russia Today, Okay Africa, and The Monthly.
  • pioneers of crowd-funded journalism like ProPublica, The Correspondent, WikiTribune, and Mother Jones.
  • and most promisingly, the newer, smaller niche publishers and bloggers that focus on specific angles. This list is long, and this is probably the future of journalism. It includes names like Jezebel, Axios, Dazed, The Ringer, Deadspin, Vox, On Being, Thrive, Gizmodo, Backchannel, Roads & Kingdoms, Vulture, PressThink, Lion’s Roar, Co.Design, WaitButWhy, FiveThirtyEight, Lawfare, Quartz, War on the Rocks, and City Lab.

There are two publishers we gave a discount; The New York Times and The Washington Post. This is not because their quality is low or because Trump dislikes them, but rather because they are overrepresented in the media picture. A primary objective of Flightmode is diversity of sources. We want to give you a wide worldview.

Then there’s a final discount in the system; the Trump discount. If anything is overrepresented in the media picture today, it’s him. We’ve set the discount at 50% so that only the very best Trump analysis makes it through.

This is not about being pro-Trump or anti-Trump. We simply believe the world will be a better place if people spend less time on Trump than they do now. We don’t think it’s good for a president to dominate our attention span the way the current one does. His media strategy seems to be to simply hog the limelight, thereby weakening democracy and undermining media. We believe the best way to ward off his media attack is by limiting his airtime.

So there will be no constant Trump updates on Flightmode. And there will only be the occasional Trump analysis if I for some reason believe that this particular article will add a fresh perspective on Trump. We don’t believe there is much media power in simply covering and berating Trump.

Onto the soft launch

So after that long pregnancy, a beta version of our baby has made it into the world. Hopefully you see why we had a prolonged and complicated pregnancy.

We didn’t anticipate Trump derailing media the way he did. More importantly, we totally underestimated the importance of filter bubbles and information overload. And most importantly, we had to dig deep to re-evaluate why people ought to spend time on journalism.

This last part is really what defined the whole journey. We moved from intellectual snobbishness (Edify), to social validation (Vouch), to societal impact (Impactable), to enjoyment (Hammock), to 100% chill enjoyment (Flightmode).

This story-work was central because the story behind media is severely flawed. The current media backdrop of scattered urgency, guilty anxiety, confused FOMO, and negative fear is the very core problem of today’s media. That’s at the essence of what we’re trying to change.

Figuring out our media story (see link), and then figuring out how to deliver that story, is what lead to this 2 year-long pregnancy.

Now that the pregnancy is over and our baby is out in the world, we want it to grow. To do that, our baby needs love and feedback. Please give it both, but most importantly, give it feedback. Our baby needs to learn. So let us know how we can improve Flightmode.

With love and listening ears,
The Flightmode team

P.S. Please e-mail anders@getflightmode.com to give us feedback.