OpenAI Wrappers
In the startup world - there’s always been that nagging question when discussing a new idea:
“What if a tech giant does this?”
Before, it was usually “What if Google does this?”
Now, the question is “What if OpenAI does this?”
You have a couple potential answers:
- They’re too slow to act on this
- They won’t focus on this particular problem
The first one depends on the company’s size and how many levels of hierarchy they have. Bigger companies tend to move slower.
The second one - you can predict with decent accuracy based on OpenAI’s main focus areas and priorities.
Some ideas will get crushed by OpenAI’s advancements. Others may actually benefit from them.
You definitely don’t want to be in the “get crushed” group.
Think of it this way:
Imagine OpenAI is creating a new programming language (like C++) - instead of AI models.
And you run a company for booking flight tickets. Your whole app is built in C++.
Would improvements to C++ kill your business? Probably not.
But why wouldn’t OpenAI just add a flight booking system to their programming language?
The main reason: OpenAI can make more money serving general use cases - like a versatile programming language. Not niche applications.
It’s not that they won’t do it. They simply can’t build every possible application.
Just like the creators of C++ didn’t (and couldn’t) develop all the money-making C++ software we have today.
The context needed for each piece of software is held by the company laser-focused on that specific problem.
Similarly, the context you have about your business - even if it’s just built on top of OpenAI’s models - is your unique application.
OpenAI is the versatile tool, like C++. Your business is a specific use-case built on that tool.
The context you provide “out-of-the-box” is your product. Products are all about convenience - like saving users from typing long prompts. People will pay for that.
For example: You could make a website that uses OpenAI to write human-sounding emails. People would probably pay for that convenience.
But some ideas directly clash with OpenAI’s path. Unfortunately, these tend to involve hard engineering problems :(
If you’re selling enterprises a C++ module for binary search - you’ll probably get replaced by the next C++ release.
Similarly, if you’re making a background removal tool or text-to-image generator - you’re probably clashing with OpenAI’s path.
In cases like that, you can only compete on output quality compared to OpenAI. Evaluate how your strategy, data, and resources stack up against OpenAI’s.
On the other hand: If you’re building an anime movie maker using Sora - OpenAI may never directly compete with you.
Creating anime requires heavily prompting the AI in very specific ways. That heavy prompting is your product.
There’s no shame in being a “wrapper” on top of AI models. We were all just wrapping C++ libraries and no one realized it.