The AI graveyard: a running list of projects and startups that didn't make it

The AI honeymoon is over. From Siri's constant delays to OpenAI's messy super app attempts, high-profile projects are failing to stick. Having a smart model is no longer enough if you can't actually execute.

For builders, it's a reality check that novelty won't save a bad product. For investors, it's a signal to stop chasing model hype and start looking at retention and defensibility.

The era of demo-driven funding is fading. Capital is shifting away from pure model plays toward companies that actually solve friction in existing workflows.

• →Build boring integration layers that fix existing workflow friction rather than trying to replace entire apps.
• →Focus on vertical AI where deep domain expertise provides a moat that general models cannot touch.
• →Prioritize distribution-first strategies where the AI is a feature of an existing habit rather than a new destination.

• →The wrapper trap, where you build features that OpenAI or Apple can sherlock overnight with a single OS update.
• →High burn rates on research-heavy projects that lack a clear path to user retention or meaningful margins.