Here's Dmitry Shapiro's big idea: apps that go inside your apps. Want to play a game inside a Reddit post, make a GIF without leaving the iMessage window, or pick theater seats with your friends? That's what Shapiro, his co-founder Sean Thielen, and their new company Koji, want to do.

The thing about that idea is that everyone else has it, too. Just last week, Apple announced "App Clips," apps that run on top of your other apps. The week before, Snap launched Minis, apps that live inside your chat window. Many others have tried over the years: Facebook Messenger has "bots" for the same purpose, and Google built the whole Allo app around the concept. The iOS App Store has a whole section of apps for iMessage.

Get daily insights from the Protocol team in your inbox

The only really successful example of a mini-app ecosystem is in China, with WeChat. There, millions of people use "mini programs" to do everything from play games to buy movie tickets to go on dates. Everything is discovered, used and paid for in WeChat. It's easy to see why it's an appealing concept and why any company would kill for the lock-in effect. Andreessen Horowitz partner Connie Chan laid out the vision by tweet in 2017, when mini-apps were again in the news: "Think of it as a better version of a browser, where you're signed in and payment-enabled for every site you visit. It's not an app store."

Many have tried to make this happen outside of WeChat, and none has succeeded. Dan Grover, a product manager at Facebook who's worked on mini apps in a number of places, said after the Snap Mini announcement that the problem was a combination of bad discoverability and a "paucity of genuinely killer use cases." Also: these apps tend to be built into chat windows, where most people, most of the time, aren't looking for anything more than conversation.

Shapiro has a history of ambitious social failures, too: He was the CTO of MySpace Music and a product manager for Google+, and he built a video player called Veoh that couldn't keep up with YouTube. In short, he has a history of being in the right arena of social media … but on the wrong team. But now he's picked a space with no winners so far. And he thinks he's got it this time.

Koji is a twist on the mini-app idea. Rather than make apps that work inside a messaging app, Shapiro wants to make apps that work everywhere. Imagine a dating app you could embed in a text thread and browse with your friend, or a Twitter clone just for you and your friends. Shapiro said the Koji team's made all that and more. "I used to do demos for VCs, where I would say, 'Let's make our own social media service. Let's remix Twitter, change the logo, change the name, change the primary and secondary colors. And let's say people can only authenticate with stanford.edu email addresses. I can do that in three minutes: push a button and launch a brand new social media service only available to Stanford people."

More likely, Shapiro said, Koji could replace all the single-purpose apps in a user's life. Blood pressure loggers, habit trackers, movie ticket buying, food ordering, boarding passes, all of it. Those rarely need to be native apps, and in fact they'd be far more useful embedded in another app you'd rather be using. "You can send it on email, you can embed it in social messengers, whatever."

Every Koji also has a QR code associated with it, which means they can be quickly accessed from the real world — another thing it shares in common with Apple's App Clips. A diner could scan a code at a restaurant and view an interactive menu while they wait for their table, or just play a themed game to kill time.

Kojis can be games, memes, or remixes of either — and they work anywhere there's a web browser.Image: Koji

That's the long-term goal, anyway. The short-term plan mostly involves silly, simple, endlessly customizable video games. Imagine a Flappy Bird clone, but instead of a bird, it was your head flying, and instead of pipes you're flying through a world made of Whoppers. (Brought to you by Burger King.) "We think that interactive is the next frontier," Shapiro said, "and not only interactive, but this kind of remixable interactive." Koji wants to go beyond pictures and video, even past TikTok's music remixes, giving users the tools to personalize and tweak every aspect of what they see and do and hear.

The start-with-games strategy isn't unique to Koji either, by the way. When Snap announced Minis, director of product Will Wu noted that they were built on the same infrastructure and user experience as Snap Games. "This foundation we built can be used for so much more," he said, before launching into a bunch of examples: meditating with friends, buying movie tickets, registering to vote, and more. "They're incredibly easy to develop, and they work for all Snapchatters on Android or iOS, with no installation required," Wu said. Sound familiar?

Here, too, Koji hopes openness makes the difference. Everything made in Koji is a web app, based on JavaScript, which means it will work everywhere. Users can embed a Koji in a tweet or a Facebook post or send one to friends in a text message. "They're not native apps, they're just web URLs," Shapiro said. "You don't need to install and there's no friction." Koji doesn't require any new apps, any new logins, anything at all. "You don't even need to know it's called Koji," he said. "You just know that inside Facebook or Instagram or Snapchat, you have new capabilities to take the post you just saw, remix it and post it again."

It's all possible thanks to standards like WebRTC and WebGL, which have made web apps capable of doing most things native apps can do. In fact, native apps are increasingly webby, as they turn to Electron and other tools to make it easier to develop for multiple platforms. Shapiro wants Koji to go one step further and ditch the apps entirely. By integrating with Stripe, Koji can enable payments and micropayments, so developers could charge small amounts for their apps or remixes. (As Shapiro points out, an awful lot of people would pay to play a game just because the main character had their favorite YouTuber's face.)

To get things started, the Koji team built a bunch of templates. You may have even seen a few, since Koji's been operating in stealth for a while now. They're mostly Flappy Bird-esque: simple graphics, easy game mechanics, often surprisingly fun. But the gameplay is only part of the point. Tweaking the levels, changing the characters, all with just a few taps and clicks, that's the real fun.

Beyond games, Koji also built interactive selfies, where users can flip between two images with a tap. There are meme templates, image-editing tools, all the simple things you might find on Instagram. Every template is easy to change, from the images to the game levels to the source code itself. Koji's hoping that not only will developers build on top of their templates, they'll build a massive library of new ones. Some free, some paid, all available virtually everywhere.

In addition to remixable games, Koji users made meme templates — like this one spoofing a Fox News screenshot.Screenshot: David Pierce/Protocol

Shapiro told me he's been thinking about this idea since he was at Google. "We started talking about progressive web apps and Instant Apps," he said, "and there were these competing teams within Google saying, 'Look, the concept of creating native apps for everything, this doesn't scale.'" He compared the native-app era to the era of buying software in boxes from Best Buy. Eventually, you could get software over the web, and it changed everything. Shapiro argues that Apple and Google are running big-box stores for apps, and the web needs to blow up the system again.

It all seems a bit silly so far. But Shapiro says that's the idea. "If you look at YouTube in the early days, it was cat videos. eBay was Pez dispensers. Amazon was books. This is our cat videos. These are fun, they're interesting, people love them. They post them, they do their thing. But in our mind, this is just .001% of what this thing is." He said when Koji first started showing developers the platform, telling them they could make anything they wanted, they didn't know where to start. Games and memes gave them a place to start.

Someday, Shapiro has ideas about building a social network inside Koji, so you could follow someone's creations and remixes. Here, too, YouTube is a good example: It grew in part because it was so easy to embed YouTube videos on other websites, but eventually people started going directly to YouTube. But for Koji that comes later. Shapiro's thinking on issues like moderation is similarly early: Koji screens for nudity and will take things down when requested or reported, but there's no system for fighting hate speech or disinformation. Even the very early content makes clear that Koji will need a system soon.

Even in these early days, though, Koji has distribution nailed. Anywhere the internet works, Koji will, too. The question going forward is whether Koji can convince developers that they should build Kojis instead of Snap Minis or iMessage apps. It's the app store versus the open web all over again. Only mini.