The Download: how people really use AI, and Flock's design choices

AI giants like OpenAI and Anthropic are gatekeeping the truth about how people use their tech. Their usage reports are curated highlights, leaving builders to guess at the real, messy workflows happening in the wild.

Building on sanitized usage data is a trap. If you're designing workflows based on what big labs say is 'popular,' you're likely missing the actual friction points where people need real help.

This information gap favors specialized startups over generalist labs. Companies that can observe and productize real-world friction will pull ahead of those just chasing model benchmarks.

• →Build vertical-specific agents that solve unrecorded tasks like data cleaning or messy triage, instead of just general chat.
• →Create observability layers that help enterprises see how employees actually use LLMs, bypassing sanitized lab reports.
• →Focus on 'invisible AI' that integrates into existing workflows rather than forcing users into new chat windows.

• →Founders chasing 'vanity use cases' that look good in demos but don't provide repeatable value for actual users.
• →Developing products based on the 'showcase' behaviors that big labs prioritize in their marketing and training.

[{"heading":"The PR Filter","body":"Big labs release usage reports that act more like marketing than science. They highlight the 'magic' moments to keep valuations high, which leaves a massive blind spot regarding how people actually use these tools in their day-to-day lives."},{"heading":"Design for the Grunt Work","body":"Flock's design philosophy suggests the win isn't in making the model smarter, but in making the interface feel inevitable. It's about moving away from the 'chat box' and toward tools that anticipate what a user needs before they even type a prompt."},{"heading":"The Intelligence Asymmetry","body":"This creates a massive opportunity for the next wave of AI startups. While the giants fight over general reasoning, the real value lives with those who understand the unrecorded, messy, and unglamorous parts of human work."},{"heading":"What to Watch","body":"Watch for a shift in how AI companies report their metrics. If we see more 'task completion' data instead of 'session length,' it's a signal that the industry is moving from hype to actual utility."}]