Risks and Rewards of Building on OpenAI

Reading Time: 5 minutes

Should you build an indie business on top of OpenAI? Absolutely. It’s not in OpenAI’s interest to build niche solutions that compete with indie hacker startups. It’s in their interest to enable them.

The community is caught up in a strange back-and-forth between “look at all these amazing opportunities” and “don’t build, OpenAI is just going to make your startup evaporate.” If you’re an indie hacker and you wonder which of these is actually helpful for your journey, I’ll try to help clear things up.

Let’s start with the elephant in the room: “wrapper.” Everyone is talking about just how short-sighted it was to build a business on wrapping OpenAI interfaces into a standalone app. And they’re right: if your business success is predicated on making it slightly cheaper to access what OpenAI offers, you’re in trouble. They can —and do— change pricing, access restrictions, and interfaces at any point. That lack of control is a severe business risk.

Experience this article as a podcast, a YouTube show, or as a newsletter:

With OpenAI’s recent release of features that threaten some businesses, such as “Chat with this PDF” kind of services, it raises the question of whether it’s a good idea to build a business on top of OpenAI.

Well, let’s look at what Damon Chen, founder of PDF.ai has to say. His business should be massively impacted by that change, if the developer community is to be believed. But it seems he’s doing just fine: “OpenAI’s updates excite developers. pdf.ai’s updates excite customers. We are not the same!”

And that’s why this whole debate is so problematic. It focuses on technical details, but completely ignores the idea of customer inertia. When customers find a product that reliably solves their problem in a very specific way, they don’t jump ship towards a more generic implementation. Particularly when your customers are not technical people. If, for a few dollars a month, I get an up-to-date version of something I would have to build, maintain, and budget for myself, I won’t even look at the platform.

Even technical customer don’t just churn from a tool that was built specifically for their purposes. Look at how many specialist hosting providers exist for web applications. If it was about saving the most money, every single person would host on AWS. But in reality, people have different preferences. They want more than a generic platform.

And that is the core aspect that I’m missing from the whole “OpenAI-kills-startups” conversation.

It is not in OpenAI’s interest to do this — particularly as these “imaginarily dead” startups wouldn’t even exist without OpenAI in the first place. As more capabilities are being added to platforms, more businesses spring up. Sure, some of them fail, but that’s more competition-related than anything else.

OpenAI can’t handle the complexity of a specific solution. Sure, they have PDF introspection. But will they be building bulk PDF imports? White-labeled PDF embeds? An externalized API to chat with a particular PDF through a REST interface. Of course, they won’t. Their core business is training and selling Large Language Models.

This is where focused solutions come into play. They take the generic API of the platform and use it to construct something tailored to the well-understood needs of professionals.

These aren’t wrappers. These are the businesses that survive changes in Open AI. Not only that: every time OpenAI makes things easier, these businesses flourish! Bhanu Teja of SiteGPT said it best: “No more spending time trying to manage embeddings, and providing context from these embeddings. We spent a ton of time building this in-house. Now every thing is about to become a simple parameter in an api call. We can delete a lot of our custom code now.”

Founders with a reasonable grasp on AI have understood that if you build a business on top of such a powerful dependency, you don’t just get ChatGPT. You get leverage. And you get data.

Think about it: if you’re helping people write Bride’s Maids speeches, starting with a few notes and ending with a customer-approved speech, what do you have? Assets! You have captured a magical transformation from idea to tangible speech, and for every customer who makes such a speech, you can add a new record to your database. Over time, you can use AI —internally— to learn what good speeches are made of. Oh, and you also have all the bad speeches that the customer didn’t like. You have insight into styles and topics that negatively resonate. This is all valuable information that no one else has. You may not control the means of production, but you own the ingredients, the recipe, and the finished product.

Meanwhile, with every new customer of your AI business, you’re building an audience around your expertise as a problem-solver. These relationships persist through any ups or downs your particular business might have. With people who sign up for your business, you control the means of communication. Every new customer adds to your reach, even if you just pass through their API calls to OpenAI. And if your product works, they will talk to their equally AI-curious peers.

Even the “wrappers” are powerful leverage generators.

How about we all calm down a little.

There’s no guarantee that your business won’t be impacted by changes within OpenAI. The platform risk is clear, and the speed of development can make your product irrelevant in a few months if you don’t offer more than just slightly easier access to OpenAI. But if you’re quick at building features and getting feedback from your users, you will use these innovative new AI tools to build special-interest startups that people have never seen before.

Look at new features like GPT-4 Turbo with its increased token limits. So many things are now possible that were once difficult or expensive. You can ingest hours worth of podcast transcripts, for example. From there, you could build products for podcasters, listeners, publishers, audio professionals, you name it. None of these people want to log into the OpenAI console and manually write a prompt. They all want well-integrated tools.

For those with low risk tolerance, consider generative AI tooling as a feature, not as the core of your product. Use OpenAI as an additional component to your offering. Many startups are using OpenAI without relying solely on it. This way, your business strategy won’t be derailed by quick changes in the AI platform.

If you have a high appetite for risk, you can build products using OpenAI tools and try to monetize them quickly. Tony Dinh did this with TypingMind. He built a model-agnostic interface that interacts with many different platforms and offers several convenience features. Will it last? Maybe. It’s just a UI for GPT, and that’s all it was ever meant to be. It’s a Buy-once-use-forever product by design.

These specialized and well-scoped tools can cater to particular needs that general-purpose tools from OpenAI can’t address. So, if you build a tool that solves a specific problem for a certain type of user, your business still has a chance to compete with the platform itself, because they need to support every kind of user.

So, should you build with, on, and around OpenAI.

As an indie hacker, I’m excited by the technology. It’s a great enabler.

We need to create tools that use AI to make our lives more interesting, easier, and improve how we learn from each other. So, if you want to build a tool today, consider OpenAI.

If you’re risk-averse, don’t build on top of it exclusively. Instead, see it as a service you can use to add value for your customers in ways you’ve already been doing without AI.

If you enjoy the wild ride of cutting-edge tech, make OpenAI your core dependency, but quickly build features that the platform will never have the incentive to create, like industry-specific data transformations or workflow integrations.

In either case, show that you want to solve a specific problem for specific people. Make them care about your product beyond the magic of GPT. That’s how your business stays alive when OpenAI announces the next big thing.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.