If you haven’t already read Packy McCormick’s blog post this morning on Attention is All You Need, I highly recommend it.
He always writes with incredible insight, but today’s post is especially important as we think about technology in the next several years and decades to come. Of course, you guessed it – his post is about OpenAI. I want to build on that post and suggest some thoughts on where the best places to build are in light of where OpenAI is positioning itself in the market.
A few months ago, we saw the launch and rise of ChatGPT, a personal assistant that allows users to ask questions and make requests. ChatGPT is the first product ever to reach 100m users in just a couple of months, an incredible feat that surprised just about everyone. Then, last week, we saw the launch of ChatGPT plugins. With the launch of plugins, users will be able to extend ChatGPT’s capabilities to be able to take actions available across other websites within the ChatGPT interface.
For example, in the future, within ChatGPT, I ought to be able to ask ChatGPT to:
- Text all my Facebook friends a fun made-up song on their respective birthdays
- Search Kayak for the cheapest premium economy flights to London from San Francisco that serves Haagen Dazs ice cream mid-flight
- Write a script for me that will re-organize my email inbox, prioritizing my founders first
Months ago, I think we saw all of this functionality coming down the pipe. But up until recently, I wouldn’t have thought that OpenAI themselves would develop a consumer-facing product to tackle all of this functionality. Like many others, I thought that OpenAI would continue extending their platform to more and more startups and developers to build on top of their technology.
What does this mean for developers?
The crux of the matter is that OpenAI is currently running two strategies right now – a platform strategy and a consumer strategy.
Examples of the platform strategy include Apple and Zapier. Without developers, Zapier would cease to exist and Apple would be largely useless without any apps on their devices. Although both have created “example apps”, neither company really attempts to build their own consumer-facing applications that compete with their partners’ products.
Examples of the consumer strategy are companies like Facebook and Twitter. You go to those sites and you generally stay on those sites, scrolling through all their content. And, although both started out with developer platform programs, at this point, both largely have just built all the functionality on their sites.
No site can be perfectly labeled a platform strategy or a consumer strategy. E.g One could argue that Apple created Numbers, which is a consumer-facing application, but let’s be real, who uses Numbers? 🙂
And in fact, running both a platform and a consumer strategy simultaneously is challenging. As Packy mentions in his post:
“All of that is if OpenAI decides to play nice with its partners. If OpenAI optimizes for its ChatGPT users, though, it’s going to disintermediate a ton of businesses and force them into changing how they operate.“
In fact, an example of a company that struggles with trying to run both a platform strategy and consumer strategy is Amazon.
Amazon started out as a platform for e-commerce stores to build on top of. I have friends who run online stores who initially hopped on Amazon and reaped the benefits of their logistics support, distribution reach, among many other helpful features.
But you may have noticed that Amazon also now sells their own end products – everything from shoes to paper shredders. And their products are GREAT. One of my friends who runs an online store found that at first, it was hugely beneficial to her business to be on Amazon, but eventually, Amazon came out with an identical clone of her products, utilizing all the data from their Amazon presence to know that she was making bank, and Amazon wanted a piece of that action.
Needless to say, my friend knew she wouldn’t win on Amazon, so she left their platform. And her store continues to grow and thrive as an independent site. In many ways, Shopify exists as the platform strategy in this space to challenge the power of Amazon and provide an alternative for e-commerce stores as Amazon adopts more and more of a consumer-facing strategy.
This brings us back to OpenAI – if there’s even an outside chance developers think OpenAI will copy them and integrate their functionality into ChatGPT, then top developers won’t want to partner with them. This creates a large opening for a potential competitor to truly go after the platform strategy. But said competitor will have to move quickly AND will need to reassure developers that they will not launch consumer-facing offerings that will compete with their own partners.
In many ways, one might be wondering how OpenAI ended up in this situation of pursuing two conflicting strategies. Pure speculation here, but I can see how one might like to start w/ the platform strategy. But, in launching ChatGPT – perhaps initially just as an example app – much like how Numbers is an example app, they may have been as surprised by their fast and impressive distribution success. And if you’re able to attract 100m users that quickly, then it only makes sense to double-down on the consumer strategy and ditch the platform strategy. Afterall, owning the end user experience and being the brand that a consumer remembers is the preferred strategy here if you can pull it off. It allows you to compete with Google and just about every destination site on the internet. It only makes sense for them to switch strategies at this point.
So who ends up serving developers?
There are certainly a lot of would-be OpenAI competitors looming around, but right now, it’s not clear to me who will move into this opening that OpenAI is creating. This will be a land grab to own the platform strategy in the AI space. Platform strategies often have one of two things that partner companies gravitate toward: 1) Distribution and 2) Technology.
For the former, we’ve talked about why people build on Apple or sell on Amazon – clearly distribution. In this case, a would-be competitor to OpenAI would have to offer strong technology.
An analogous example that comes to mind here is TSMC. TSMC is arguably the best contract chip designer and manufacturer in the world. Although most people think that being in the semiconductor chip business is a horrible commoditized business, TSMC has margins upwards of near 50%! If you’re competing on technology utilizing a platform-strategy, you’d better be #1 or perhaps #2 in technology, because if you are #8, your tech largely is a commodity.
TSMC manages to generate these kinds of margins, because they can produce very tiny chips, which means that phones and computers can cram in more chips into a limited amount of space for more computing power for their devices. Unlike in a consumer-facing strategy, no one cares if your chips are more user friendly or look nice or you have better customer support. The only thing that matters with this strategy is that you have superior technical specs. TSMC has continued to be the market leader over the decades, because they continue to reinvest vast amounts of profits and capital into staying the leader in this market.
This is what it will take to win the platform strategy in AI. Strong technology is a starting point but continued betterment is equally important to maintain the lead. And in AI, how do you come out with better and better models? By amassing more and more data.
So I think what we may end up seeing here are weird alliances – I could imagine some sort of partnership between websites that have a lot of data and an OpenAI competitor here so that existing large destination sites will not lose to ChatGPT.
In fact, we’ve seen this movie before. When Apple launched the iPhone, Google rallied around the Open Handset Alliance as a hedge to ensure that Apple would not soak up all the mobile traffic in the world. With this open working group consisting of top mobile carriers and handset makers to support Android, an open source mobile operating system, Google was able to come out with a competitor that they wouldn’t have been able to develop on their own. In a similar vein, when social networks were taking off and Facebook was dominating, Google rallied around Open Social, which was an open source social app SDK with support from many of the top social networking competitors to try to compete with Facebook at that time.
We will definitely see something similar here – whether it’s through the alliance of various existing players to compete directly with ChatGPT’s vision or a single player attempting to go head-to-head with OpenAI on the developer-strategy much like what Shopify did with Amazon. In the ChatGPT world, you will be able to do everything on their site and in a competing world, you will be able to do much more on all the other sites. Both versions will win, because power is never left unchecked.
So what should I build?
With the AI world in flux, how should I, as an entrepreneur, navigate this? I think that we should assume that every site in the future will utilize AI to provide a better user experience with greater personalization.
So if you’re trying to build the next AI image creator or the AI trip planner, etc, consider it too competitive already. In fact, Adobe, Canva, etc are all rolling out their generative AI tools within their own existing programs. Even for websites that have terrible tech teams, you should assume there will be an OpenAI competitor who will serve those tech-unsavvy websites to make adding AI to any website really easy.
In software, distribution often wins, and as such, if I were building a new product, just like at most other points in software history, I would go after the spaces where I either a) had distribution edge already, b) saw an opening where there’s not an existing large player with distribution, or c) had an insight into a uniquely differentiated product in an existing space. And I might partner with an OpenAI competitor that is friendly to developers to ensure OpenAI doesn’t integrate my functionality into ChatGPT.
And if you are building infrastructure in AI, I think the opportunity is now to go after a developer-focused OpenAI competitor. But you need to be #1 and stay #1.
Sign up to read more posts like this one: