The Specter of Platform Risk
To understand the anxiety, you need to understand “platform risk.” It’s a classic Silicon Valley nightmare. Imagine you’ve built a brilliant app that does one thing incredibly well on, say, Apple’s iOS.
Your business thrives. Then, at its annual conference, Apple announces a new, free, built-in feature that does exactly what your app does. This is called getting “Sherlocked,” named after Apple's Sherlock search tool, which began incorporating features from popular third-party apps, rendering them obsolete. The platform you built your castle on just became your conqueror. This isn’t necessarily malicious; it’s the natural evolution of a powerful platform. The provider of the core technology (the OS, the social network, the AI model) will always be tempted to absorb the most popular and useful functions happening on its turf. For the smaller developer, it’s a constant, existential threat. You’re living on borrowed land, and the landlord can change the terms—or reclaim the property—at any moment.
OpenAI: The Ultimate High-Wire Act
What makes OpenAI a unique and terrifying source of platform risk is its speed and centrality. Unlike Apple, which updates its OS annually, OpenAI pushes foundational changes at a dizzying pace. The core product—the AI model itself—is the “platform.” Thousands of startups are built as thin layers, or “wrappers,” on top of OpenAI’s technology. They use its API to power everything from AI writing assistants and code generators to customer service chatbots and voice transcription services. This is where the headline quote, “This changes the stack,” comes in. In developer parlance, a “stack” is the set of technologies used to build an application. When OpenAI releases a new model like GPT-4o, it’s not just a minor update; it fundamentally alters the capabilities of the base layer. A feature that was once difficult and expensive to build—requiring a separate startup to solve—can suddenly become a single, simple command. The ground beneath the entire ecosystem shifts.
The Kill Zone: When Features Become Landmines
The May 2024 unveiling of GPT-4o was a masterclass in this phenomenon. OpenAI demonstrated a model with stunningly fast, emotionally nuanced voice capabilities, the ability to see and interpret the world through a phone’s camera, and real-time translation. The demos were magical. They also represented a direct threat to entire categories of startups. Companies that had raised millions to build AI conversation companions? Their core value proposition was now a default feature. Apps that specialized in analyzing data from PDFs or creating charts? GPT-4o could do it natively. The vibrant ecosystem of “wrapper” apps, which essentially put a nice user interface on a specific OpenAI function, was suddenly on notice. Why pay for a subscription to a separate service when the core platform you already use does it for free or as part of a bundle? In an instant, OpenAI’s new feature list became a kill list for many young companies.
The Developer's Dilemma
This creates a profound dilemma for developers and investors. The opportunity is too big to ignore; you *have* to build with AI, and OpenAI provides the most powerful and accessible tools. But if you build something too obvious or too dependent on one specific capability, you risk being Sherlocked by the next model update. The smart money is now on finding a “moat”—a competitive advantage that the platform can’t easily replicate. This might mean focusing on a highly specific, niche industry where domain expertise is key. It could involve proprietary data that OpenAI doesn’t have access to. Or it could be about building a unique brand, user experience, and community that people love, regardless of the underlying tech. The strategy is to use the platform as a component, not as the entire foundation. The goal is to build a business that is *enhanced* by OpenAI, not one that is simply a feature of it.
