OpenAI Is Just Another Boring, Desperate AI Startup (www.wheresyoured.at)

A blistering critique argues that OpenAI has drifted from being a focused foundation-model lab into a sprawling, hype-driven startup leaking plans for everything from Sora 2 (a generative-video social feed) to chips, consumer hardware, an ads business, a hiring platform and a browser. The piece says GPT-5 underwhelmed and costs more to operate than prior versions because of how it processes prompts; API sales are a tiny share of revenue while ChatGPT subscriptions (reportedly ~20M paid users and ~5M business seats) still dominate. The author cites financial strain—H1 2025 costs of $6.7B versus $4.3B in revenue, heavy R&D burn against ~$17.5B cash—and estimates Sora video generation costs roughly $5 per clip using Azure rates. Agents and new products are described as “barely functional.” For the AI/ML community this raises three key implications: strategic risk from trying to be “everything” (consumer apps, cloud, hardware, ads) instead of advancing model quality; economic pressure as generative AI productization proves costly and commoditizing; and technical limits that remain unsolved—most notably LLM hallucinations, which the author says are mathematically unavoidable. In short, the story frames OpenAI not as an inevitable industry leader but as a well-funded startup scrambling for monetization while facing product, cost, and foundational-model limitations.