What is a 'Wrapper,' Anyway?
Imagine you’re selling bottled water. You don’t own the spring; you just buy water from the city, put it in a really cool-looking bottle, and market it as a premium experience. That’s essentially a “wrapper” in the tech world. In the OpenAI ecosystem, countless startups emerged by taking OpenAI’s powerful but raw technology—accessible via an API (Application Programming Interface)—and putting a user-friendly interface or a niche workflow “around” it. For a fee, these companies would help you “chat” with your PDF documents, summarize long articles, or generate marketing copy, all using OpenAI’s engine under the hood. They were, in effect, fancy bottles for OpenAI’s water. The business model was simple and, for a time, incredibly lucrative.
The Day the Features Dropped
The
nightmare scenario crystalized during OpenAI’s first developer conference in November 2023. In a series of blockbuster announcements, the company unveiled new features that directly replicated the core functions of hundreds of these wrapper startups. They announced GPTs, which let any paying user create custom chatbots for specific tasks—like analyzing a PDF—with no code required. They released an “Assistants API” that made it drastically easier for developers to build complex, stateful AI agents. They massively increased the “context window,” allowing the AI to remember much more information, a feature many startups had built complex workarounds to achieve. For founders watching, it was like the city had decided to start selling its own fancy bottled water for free, right next to their kiosk.
'Getting Sherlocked': An Old Story
This painful experience isn't new; it has a name in Silicon Valley: “Getting Sherlocked.” The term dates back to the early 2000s, when a popular third-party search tool for macOS called Sherlock was rendered obsolete overnight after Apple released its own, very similar search feature and named it Spotlight. It’s a classic tale of platform risk. When you build your business on someone else’s platform—be it Apple’s App Store, Facebook, or now OpenAI—you are living on borrowed land. The platform provides the customers and the tools, but it also dictates the rules. And the unwritten rule is that the platform owner can and will eventually absorb the most popular and profitable functions into its own core product. They see what you’re building, see that customers want it, and have every incentive to build it themselves to make their platform stronger.
The Platform Always Wins
From OpenAI’s perspective, this isn’t malicious; it’s logical. Their goal is to make their technology as powerful and accessible as possible. They watch how developers use their tools, identify the most common and valuable use cases, and then simplify them. The wrapper startups serve as a form of free market research, illuminating the path for the platform’s own product roadmap. This dynamic creates a brutal paradox for entrepreneurs. The easiest way to get started in AI is to build on top of an existing model. Yet that’s also the most vulnerable position to be in. The very thing that makes your product useful—its proximity to the core platform—is what makes it so easy to replace.
Is Building on OpenAI a Lost Cause?
Not necessarily. But the era of the simple wrapper is likely over. The “nightmare” is a wake-up call for AI startups to build a more durable competitive advantage, often called a “moat.” Instead of just repackaging a generic API, successful companies will need to focus on what the platform can’t easily replicate. This might mean owning a proprietary dataset, integrating deeply into a specific industry’s complex workflow (like legal or medical compliance), building a strong brand and community, or delivering a user experience that is ten times better than the platform’s default offering. The startups that survive and thrive won't be the ones selling the bottled water; they'll be the ones who build an entire irrigation system for a specific, high-value farm.
