- Short-Form Content, Authored by Anyone, Strung together for Consumers by an on-the-fly Algorithm
- Three brief histories
- Lessons
- To Wrap it Up
Short-Form Content, Authored by Anyone, Strung together for Consumers by an on-the-fly Algorithm
Back in 2013, I was involved with three different EdTech startups that were all trying to build a certain kind of product. Here are some of the distinguishing features we were working on:
- The content should be super short, in fact as short as it can possibly be and still make sense
- Let anyone in the world easily add their own content to the platform
- Our algorithm would monitor the consumers of the content and make intelligent choices on the fly about exactly which piece of content to show them next
The theory behind building such a company is that people all over the world would be drawn to contributing content because it would be so easy, and because it’s a lot less work to add short content than long content. Also, our algorithm would do a great job stitching content together into a great experience for the consumers because it would quickly gather data on which morsels of content were the best. By keeping the content as short as possible, we’d have the tightest possible feedback loop between showing someone a piece of content and identifying whether or not it was any good.
Does all this sound familiar? It hit me just recently that we were all nearly building TikTok back in 2013. But we did not in fact build TikTok. We built Edagog, Outlearn and Knewton, and none of them ultimately achieved its goals. Upon some reflection in 2024, the big mistake behind each of these was that they were all “education” companies or “learning” companies. They should have been “entertainment” companies. Our metric for whether the content was any good was always going to be some kind of learning assessment. Multiple choice questions, or a graded essay, or something similarly scholastic. It should have been the amount of time our users spent engaging with the content…. that metric is much easier to get, a much faster signal, and much more pleasant for the end user to give to you. They key ideas around short-form content, algorithms deciding what to show you next, and user-generated content were all there, but the whole education angle was what ruined it.
Three brief histories
Edagog
Edagog was a project by Neil Ratna, Rob Hass and me. We built a simple prototype content platform, never got any good content into it, never got any real users, and decided to go our separate ways before we were able to get any real funding or interest. We didn’t focus on video content to start, just text. We were also very education focused, and every sequence of content shown to a user was guaranteed to pepper them with multiple-choice or short-essay questions.
Outlearn
Outlearn was an EdTech startup in Boston, run on a 2 million seed investment by General Catalyst. We built a usable product, signed up a very small number of companies to try it out, had a small number of very unenthusiastic users and were eventually sold for parts to Google. One notable choice by Outlearn was to market itself to L&D departments at companies with teams of 40 or more software developers. The theory was that learning was most lucrative if it helped high salaried people improve their most lucrative skills (did I mention we were Venture backed?). It turned out that L&D departments were not all that keen on paying for our platform. Also, we initially launched with a lot of repurposed content. It all came from elsewhere around the web, and we sort of just reassembled it into our own “learning paths.” We didn’t have any content that was at all original, and none of it was designed for our platform. One of the lessons here is that great content is essential for a product like this. TikTok the platform is nothing without its content, and the content is authored specifically for TikTok.
Knewton
Disclaimer, I never actually worked for Knewton, but a friend did, and I went to visit him and the rest of the team in NYC. I got to know them, heard the pitch, told them my thoughts and had pretty high hopes for their product. Knewton was also working on short-form-content with an algorithm stringing it together for the end users. Again, it was strictly an “education” company. They ultimately created some courseware for Arizona State University. This company made by far the most progress of any of the three, brought in a lot more funding, and had large numbers of people using the platform. They were never up for letting just anyone on the internet create content. Instead, they wanted the content to be created by subject matter experts. Knewton was not for fun, it was for school.
Lessons
Today, I regularly encounter young people that seriously rely on TikTok to learn about the exact same topics I remember hoping to have on Edagog/Outlearn/Knewton. People use it to study politics, to learn history, to get their news…. and of course it’s vastly more popular than any of our EdTech products ever hoped to be. It struck me the other day that TikTok is already the real-world actualization of the goal of the EdTech companies we worked on 10 years ago. We shouldn’t have been building school, we should have been building something more like TikTok. There are a few lessons to this story:
Lock in the Essence and Leave the Rest up to whatever Works the Best
For any startup idea, it helps lock in the “essence” of the idea, and be extremely willing to change anything else about it. The essence here is minimally short-form content, authored by anyone in the world, and strung together for content consumers by an intelligent algorithm that makes on-the-fly choices about the best next content item to display. Everything to do with education, L&D, and multiple-choice/essay questions was peripheral. If only we’d thought about our product that way, we would have made much better choices. In particular, it would have been clear that we’d be much more popular if our algorithm used the time spent engaging with the content as the metric instead of constantly pestering our users with quiz questions.
Entertainment First, Education Later
There’s a saying that “you get no vitamins from the broccoli you don’t eat.” Applied to this case, it means that you’ll never educate people if they’re not first eager to use your product. TikTok was built as entertainment first. Even now, you could still argue it’s more of a vapid time-sink than a promising education tool. Either way, I think its success indicates that anyone trying to build a TikTok-for-education should attend first and foremost to making their product fun and engaging, and only when it’s clear that the users actually want to use it should they get around to optimizing for learning objectives. Which company should build TikTok-for-education? Probably ByteDance. Perhaps TikTok already is TikTok-for-education, but that doesn’t mean other imitators can’t still give it a shot.
Big Successful Ideas are (almost) Obvious
A lot of very successful company ideas seem kind of obvious in retrospect. They can also seem kind of obvious in advance, but often, people who are barking up the right tree are still missing some key ingredient. The core idea behind TikTok’s success was (almost) obvious to us all in advance, but we were so locked into thinking of specifically short-form educational content that we didn’t realize the right category was short-form content generally. So we almost had the same idea, but really we didn’t.
Another fun example, one of my coworkers at Outlearn was the serial entrepreneur Will Koffel. Before working with us, he’d had this great idea that you should be able to order a taxi using an app on your cell-phone. He even formed a company working to do exactly that, but it launched, and ran, and fell apart all before the iPhone showed up. The general idea was that an improvement to the User Experience of ordering a taxi would be super impactful for the end-user, lead to a very successful company, and that advances in cell-phone software enabled exactly the kind of User Experience you needed. The idea was (almost) correct, but the missing piece was that the pre-iPhone UX wasn’t quite good enough, and you had to wait just a little longer for phones to be ready to support the kind of UX you needed.
To Wrap it Up
We had high hopes for these three EdTech companies, but all of them should really have built something much more like TikTok if they’d been smarter about achieving their goals. They key idea behind TikTok was also there at each of these companies; we wanted to build a platform for short-form content that was easy to author, and we would use an on-the-fly algorithm to string it together for end users. None of these companies really got where they wanted to go, and upon reflection 10 years later, there are some clear and apparent reasons why.