The mobile S-curve ends, and the AI S-curve begins
Why mobile apps are meh, and AI apps are fire
You may have noticed that Tech Twitter is obsessed with AI, and just meh about mobile.
There’s never been a bigger contrast between mobile and AI — it’s the end of one technology curve, and the start of the other. It’s been 15 years since the App Store was launched; while the generative AI revolution started merely 18 months ago. Mobile is now dominated by a duopoly of two giants, administering a collection of <100 apps that never seem to leave the charts (and a long tail that doesn’t matter much). This duopoly’s only true opponents are world governments.
The brand new AI ecosystem, on the other hand, is in a state of utter chaos with new startups, technologies, and papers launched every week. One generative AI startup looks to be in the lead one week, and a few weeks later, their entire approach is in crisis — just look at AI video, recently! New approaches like open source, new types of hardware, regulation, and much more threaten to upend the stack rank every few quarters — isn’t that existing!?
Every startup is building on top of an existing S-curve. Perhaps you’re building in mobile, which is near the tail end of the curve. Or perhaps you’re working on spatial computing, web3, AI, or something near the beginning. This will color your product approach, how they take their startup to market, how investors think about funding in the sector, and so on.
THE "IT WORKS" FEATURE
When startups are early to an S-curve, as we’re seeing with AI right now, the act of building a product is often hard, but the growth can be easy. In the early days of the web, developers did not have the benefit of open source, cloud computing, high-level programming languages, and on. The sheer act of building something like eBay (something that we would now consider relatively simple), was a miracle in itself. The most important feature that a website needed to have was the “It Actually Works” feature. That is, the technology was difficult enough and the developer base small enough that delivering an actual working product was enough to attract users without much marketing or growth. You might know that today's AI landscape is perhaps not that different. You’ve now seen new startups posting a demo video or some test output and immediately there are thousands of users on their waitlist ready to try their product. In the age of social media, I think this has been accelerated for AI products that generate visual output — whether video, photos, 3D assets or otherwise — because it’s so easily shareable.
In the early days of mobile, it was often said that apps competed with boredom. Apps competed with waiting in line, sitting on the toilet, and all the other boring bits of time we’d rather be doing something else. But 15 years later, a new app has to compete with the most engaging experiences ever built — whether that’s an infinite stream of short videos, or an endless scroll of beautiful travel photos, or something else. As mobile hits the plateau of its S-curve, it’s not enough to simply “work.” In fact, mobile apps, have to be very very different than anything that has come before it to have a chance of success. Early S-curve products can fast follow and exercise Steve Jobs’s famous quote “good artists copy, great artists steal.” Late S-curve products have to contend with significantly higher user expectations of what constitutes a minimum viable product. Late S-curve startups are better off trying to create new categories, rather than fast-following, because at least new product categories might invent a small minimum viable product.
THE NOVELTY EFFECT
Novelty fuels early S-curve products. This desire for novelty means that whenever there are major new features or upgrades to underlying AI models, you see a rush of new users without much marketing effort. While all of this novelty drives high-level growth numbers, I will also argue that in the early phase of an S-curve, there is both high growth and high churn. During this phase, growth is easy because of word-of-mouth from highly enthusiastic early adopters. However, I would warn you that eventually low retention will catch up to even the fastest growing products as novelty effects start to die down.
Novelty effects kick in because, well, humans are kind of dopamine fiends. When we see an amazing AI-generated image for the first time, it’s like WOWWOWOW — followed by sharing, commenting, and forwarding to friends. But do show people a cool AI photo a few dozen times, and we get used to it. We need even more, to be entertained. Thus, sharing goes down, and so down overall engagement. Every product on an S-curve in a rapid ascent up the middle of the S, but as the plateau starts to hit, then the novelty effects start to unwind. At some point, retention will be king, and only the most retentive products will survive.
Investors focus on the early part of the S-curve because the first few years should usher in big upticks simply by being in the market. It’s easier to deliver an “it works” feature than to compete in red oceans of established products, and founders can get new users, plenty of growth, while trading off the fact it might be a big harder to build a v1. This is why venture capital money often pours into a new sector like AI, causing many startups to follow the money and pivot into the category. Is this opportunistic? Yes. Is it smart? Probably also yes. The best markets often have tons of competition, are hot and very dynamic, and this is often seen as a good thing. If you’re in a market by yourself with no competition, then perhaps it’s not that great of a market after all?
THE LATE S-CURVE
The skills needed to succeed in a late S-curve market, in contrast, are very different. And they translate to a different set of dynamics that investors will evaluate.
The product thesis has to be along the lines of “It’s radically different” than “It works.” After 15 years the mobile apps market has had, literally, millions of different forms of experimentation. Every photo app that people could have imagined during this time has probably been tried. Radical new innovations are needed, or at least innovations that are counter to dominant narratives. (An example here is the “anti-Instagram thesis” that products like BeReal have been able to ride lately)
If an older market no longer from a novelty effect, then growth will be slower and more efficient from day one. Often the growth channels that dominate at this point, our saturated by establish players. For instance, this is true in mobile where mobile ads are saturated by gaming/travel/ecommerce companies. Startups often find it hard to compete in the saturated channels, and instead have to discover novel ones which is difficult. Investors tend to seek efficient growth, rather than expecting, or seeing wildly exponential curves.
WHEN DESIGN RULES
This doesn't mean that late S-curve startups aren’t possible, or that they can’t be successful. This is where design-oriented teams can thrive. It's been noted that while Apple didn't invent the GUI operating system, the MP3 player, the laptop, the smartphone, they arrived at the middle/end of the S-curve and often perfected the product. This is last-mover advantage, rather than first-mover. But it’s a very different type of team that does this versus the ones that create the first practical instantiation of something.
Today, founders find themselves at a crossroads. Those who have hard-earned secrets in mobile apps will still find opportunities to use those secrets, but need to be mindful of the stage of S-curve they reside. They will have to be even more clever, do even more what’s new, in order to break out. And for those who stand at the beginning of the AI revolution, it’s easy to get started. But it’s hard to ride the chaos and stay on top. And even harder when the novelty inevitably stops, as only the best products will win.
Final thought - I leave you with a chart of many of the technologies that have emerged, ascended rapidly, and then slowed down over the past few years. You'll notice that as modern communication/publishing/computing has taken hold, technology adoption is getting faster and faster. Get ready to ride the S-curve.
What a great article.
It will be interesting to see how both mobile + AI intersect and whether this will revitalize the interest in mobile apps.
I know my comment is a bit, late but a good article. Just a few of more notes:
1. Not only for VC, but for everybody, technology S-curves should be in your vernacular. The valley tends to cycle through buzz words and phases, but s-curves should never be forgotten.
2. The theory for S-Curves was by Everett Rogers, and if you never read his book Diffusion of Innovation, it is a profound revelation. However, I think of a practical application viewpoint, Geoffrey Moore crossing the chasm concept has never dropped in relevance. Although it seemed it hit it's peak of popularity in the valley about 20 years ago.
3. I think you cover this in your post, but I think the concept of the chasm is incredibly insightful, and creates a vernacular to sharpen the points that you already bring up, and even more importantly put some boundaries around when a market turns real. I am attracted to the idea that the chasm exists somewhere around 10% of the targeted TAM.
The beauty of this framework is almost all business cases start off with somebody pointing out a massive TAM, and saying "if I only get 1%, we'll roll in the money." If you then say, "but I don't consider you real until you pass 10% of the TAM," it places a feedback mechanism. The bigger they call the TAM, the further they are away from crossing the chasm. The smaller they call the TAM, the less compelling their business case becomes. Identifying the chasm is a control mechanism to prevent a crazy ROI business case.