The Platform Problem
Imagine building a beautiful house on land you don't own. For thousands of AI startups, OpenAI's platform is that borrowed land. They use its powerful APIs—the bridges that let software talk to each other—to create innovative products, from AI-powered
writing assistants to complex data analysis tools. This symbiotic relationship fuels a vibrant ecosystem. The problem is, the landlord can change the rules at any time. This is 'platform risk,' and it's a foundational vulnerability for any company building its core business on another's proprietary technology, a lesson many learned during the rise of Facebook and Twitter. With AI, the stakes are even higher.
Welcome to the 'Sherlock' Zone
In the tech world, 'getting Sherlocked' is a well-known phenomenon, named after an Apple search tool that famously replicated the features of a third-party app called Watson, rendering it obsolete. At DevDay, this happens in real-time. When OpenAI announces it's adding a new, native feature—like transcribing meetings or analyzing PDFs directly within its own products—it can instantly vaporize the market for startups whose entire business was built on providing that exact service. Suddenly, a feature that was a standalone company becomes a free bullet point in a press release. This forces startups into a constant state of anxiety, wondering if their core product is on OpenAI's internal roadmap.
The Chilling Effect on Investment
Venture capitalists, the investors who fund early-stage companies, are paying close attention. The fear of a startup being Sherlocked makes them cautious about funding companies that look like 'thin wrappers'—essentially just a pretty user interface over an OpenAI API. Investors are now asking founders the tough question: 'What is your moat? What prevents OpenAI from building this themselves in a weekend?' This has led to a significant shift in funding strategy. VCs are increasingly wary of generic AI tools and are instead looking for companies with unique advantages, such as proprietary data, deep integration into specific enterprise workflows, or a strong brand community that can't be easily replicated.
How to Survive the Test
Despite the risks, not all startups are doomed. The stress test separates the fleeting features from the durable companies. The survivors are those that offer more than just access to AI. They solve a specific, niche problem for a dedicated customer base so well that a general-purpose tool like ChatGPT can't compete. They build what are sometimes called 'thick wrappers' by deeply embedding their tool into a customer's daily operations, owning the entire workflow from start to finish. Success stories often involve creating a strong data flywheel, where the product gets better with each new user, or by building a powerful distribution channel that OpenAI can't access. The goal is to build something that customers rely on for its unique value, not just for the AI model powering it.
