The article critiques the current trend of startups relying heavily on existing AI models like ChatGPT. Many companies are building agents—essentially chatbots—to solve specific problems like scheduling meetings or creating marketing copy. However, these agents often struggle due to limitations in the foundational models they use. The author points out that while these startups aim to differentiate through clever prompts, they face the risk of redundancy as AI models become more advanced and capable of performing tasks independently. To improve their agents, startups are focusing on providing more context by integrating with various data sources, such as emails and project management tools. This approach aims to enhance the quality and relevance of the AI’s outputs. However, the article suggests a potential pivot: rather than just being a point solution, companies could leverage their data integration capabilities to become context providers for other agents, essentially becoming platforms themselves. The article presents two contrasting theories regarding the value of such platforms. One theory posits that integrating data is highly valuable and that the rush to establish these platforms is justified. The alternative theory questions the conventional wisdom that aggregators are the ultimate winners in tech. Instead, it argues that the real power lies with tools that learn and push new intelligence back into the ecosystem, not just those that aggregate information. The discussion highlights a complex interplay between producers and consumers in the tech landscape, suggesting that staying focused on production, rather than trying to become a platform, may ultimately be more beneficial for some companies.