The Spring Update Shockwave
On May 13, 2024, OpenAI held a live-streamed event to showcase its latest advancements. The star was GPT-4o (“o” for “omni”), a new flagship model that seamlessly integrates text, audio, and vision. In
a series of stunning demos, the AI acted as a real-time translator, a seeing-eye assistant for a blind user, and a friendly, charismatic voice tutor. The new model was also faster and, crucially, made available for free to all ChatGPT users. For the average person, it was an exciting glimpse into an AI-powered future. But for founders and venture capitalists in the AI space, the presentation felt less like a product launch and more like an earthquake, with their own companies standing at the epicenter.
Welcome to Being 'Sherlocked'
In the tech world, there’s a term for this specific kind of existential dread: “getting Sherlocked.” The name comes from a 2000s-era Mac utility called Watson that provided quick search results. It was so popular that Apple eventually built its own, nearly identical version called Sherlock directly into the Mac OS, effectively killing Watson overnight. To be Sherlocked is to have the platform you build on (like Apple’s iOS or, in this case, OpenAI’s GPT) absorb your product’s key feature into its own core offering. Suddenly, the thing you were charging for is now a free, native part of the system everyone already uses. The panic behind the OpenAI updates is a mass-Sherlocking event. Startups that spent months and millions building a slightly better interface or a specific workflow on top of GPT technology watched as OpenAI casually rolled out their core value proposition as a free feature.
The Startups on the Chopping Block
The list of potentially affected startups is long. Any company whose primary offering was a slick voice interface for ChatGPT is now competing with OpenAI’s own, deeply integrated version. Businesses selling AI-powered meeting transcription and summarization tools saw a future where ChatGPT can just listen in and do it for them. Educational apps that use AI for tutoring now have to contend with a free, hyper-advanced tutor that can see a student’s worksheet through a phone camera. Even hardware makers like Humane and Rabbit, which pitched dedicated AI devices, look vulnerable when a free app on the phone you already own can do much of the same. The common thread is that these companies built what’s known as a “thin wrapper”—a business model that relies more on the underlying platform’s magic than on its own unique technology, data, or customer relationships.
The New Rules for AI Survival
This isn't just about OpenAI being a ruthless competitor; it’s a fundamental lesson about building on someone else’s platform. For years, startups have flocked to build on OpenAI’s API, leveraging its power to get to market quickly. It was a symbiotic relationship: OpenAI got distribution and a vibrant ecosystem, while startups got a massive technological head start. But the platform always holds the ultimate power. When the platform owner decides to expand, anyone standing on that new ground gets displaced. The panic is forcing a painful but necessary evolution in the AI startup space. Founders are realizing that a clever prompt and a nice user interface are no longer a defensible business. Survival now depends on building something truly proprietary—a unique dataset, a specialized workflow for a niche industry, or deep enterprise integrations that OpenAI is unlikely to ever build itself. The gold rush of building simple wrappers around GPT is over.
