The Platform's Long Shadow
To understand the anxiety, you first have to understand the dynamic. Thousands of startups aren't just building 'with' AI; they're building directly on top of OpenAI's platform. They use its APIs as the engine for their products, creating specialized
tools for everything from summarizing legal documents to generating marketing copy. These are often called 'wrapper' startups. They find a niche that the core ChatGPT product doesn't serve perfectly and build a business to fill that gap. The problem is that the platform owner—OpenAI—is always watching, learning, and expanding. What is a feature gap one year can become a native function the next. This phenomenon has a name in the tech world: getting 'Sherlocked'.
From Niche Product to Native Feature
The term 'Sherlocking' dates back to the early 2000s when Apple released a file-searching tool called Sherlock. A third-party developer created a complementary app called Watson that added powerful web-searching capabilities. Apple saw its success and simply incorporated all of Watson's key functions into the next version of Sherlock, effectively killing the third-party app overnight. Today's AI founders face a supercharged version of this risk. A startup might spend millions and months building a tool that lets users chat with their PDFs. Then, at DevDay, OpenAI announces that PDF uploads are now a standard feature in ChatGPT Plus. In an instant, the startup's unique selling proposition vanishes. One founder whose company was impacted by such a move called them a 'victim' of the update.
The Speed of Oblivion
What amplifies the anxiety is the sheer speed and scale of OpenAI's product launches. A single DevDay can bring dozens of new announcements. Features that once supported entire companies—like creating custom AI assistants, advanced data analysis, or agent orchestration—are now often integrated directly into OpenAI's offerings. Recent DevDays have showcased 'always-on' agents that can manage tasks proactively, a direct challenge to startups selling AI-powered assistants and automation tools. When OpenAI announces not just a new feature but also a 90% price cut on a core model, it can wipe out any competitor whose entire business model was 'same quality, lower price'. This leaves founders in a constant state of unease, wondering if their product is on an internal OpenAI roadmap.
The Investor and Talent Chill
This platform risk doesn't just affect morale; it impacts funding and talent. Venture capitalists are becoming increasingly wary of funding simple 'wrapper' startups, knowing they are one product launch away from obsolescence. Astute investors now ask a critical question: 'What is your moat? Why can't OpenAI just build this?' If a startup's only advantage is a clever prompt or a thin user interface over an OpenAI model, securing capital becomes incredibly difficult. This anxiety trickles down to employees. Working at a startup is already a high-stress gamble. Adding the existential threat of being 'Sherlocked' makes it even harder to attract and retain top talent who might prefer the stability of working for the platform itself.
Pivoting Under Pressure
So, how do founders cope? The smartest ones are learning to navigate this new terrain by building defensible businesses that AI can't easily replicate. Instead of just wrapping an API, they are focusing on proprietary data, unique workflows, or deep industry-specific integrations that go beyond what a general-purpose model can offer. They treat AI as a component, not the entire business model. The goal is to build a product where, even if OpenAI were to launch a similar feature, customers would stick around because of the startup's unique value. It’s a shift from renting intelligence to owning a problem. This means creating end-to-end solutions where the AI element is just one piece of a much larger, more complex puzzle.
