Do startups building on OpenAI have a defensible moat?

They’re a PM’s whim away from being steamrolled by their foundation model vendor. And these startups are doing the product market fit for them

The bigger the opportunity, the more general the audience, the juicier the B2B market, the greater the risk.

Time was when Jasper was valued more highly than OpenAI. Now it’s flagging and OpenAI is headed for 12 digits…

The moat could have nothing to do with the AI part. Examples:

- e-commerce app is a thin ChatGPT wrapper for the shopping experience but their moat is their products and supply chain details

- video game using LLM NPCs and other dynamic experiences but their moat is something else like incredible graphics or being based on some known IP

- any business based on curation or archival of past LLM responses or interactions, moat could be some quality of that particular data, like other data businesses

As you were submitting this question I was finishing this on-topic blog post. https://blog.waleson.com/2024/05/the-long-long-tail-of-ai-ap...

Short answer: Yes I think so.

Maybe not in the traditional sense of a moat, but maybe that's OK. It's difficult to create a deep moat with the commoditisation of technology and higher barriers to entry.

Soon enough every company will be an AI company. So to create a moat it could be a case of creating a company that wasn't possible before AI that has the characteristics to create a moat - imagine something like a personal cooking assistant that links up with your health apps and buys your groceries for you, optimised for your health - add a yearly subscription for next-day delivery and you have your moat.

Depends on whether the startup is building a feature or a product.

A lot of AI startups that are building on top of OpenAI are effectively building a feature, a small very specific solution for a small use-case. Most of those are going to fail when OpenAI or some other entity with a product will add that feature to their product.

But, if you solve a real large pain point that has value and use LLMs via OpenAI to add 1 or more features to your offering which allows it to solve that pain point more effectively and differentiate from competitors, then you have a moat.

OpenAI and Anthropic builds extremely fast, even more so than most startups. There's little point building things like langchain, because OpenAI will build a better product a few months later.

But you can build a product connected to OpenAI fine. Anthropic uses a similar API. Most of the smaller LLMs will also copy the same design so switching is cheap.

It's probably not a smart idea to be completely dependent on OpenAI either. They're down all the time, can ban you on a whim, and their customer service is inaccessible.

does people build using javascrapt or python or java programming language have a defensible moat since everyone can use that launage and immiate your ui/ux ideas.