Imagine handing over the keys to your vacation home and the controls to your irrigation system to an AI agent. No human supervision, just a set of goals and a massive amount of data. For Javier de la Torre, this wasn't a thought experiment—it was a reality that resulted in one of the most expensive heads of lettuce in history.
From Enterprise Data to Backyard Dirt
Usually, when techies talk about a "lakehouse," they aren't talking about a cozy cottage by the water. They're referring to the architectural blend of data lakes and data warehouses—a "control plane" that Databricks and Microsoft have been pushing for enterprise AI agents. But de la Torre took this corporate concept and applied it to the physical world. By giving an agent a lakehouse (the data architecture) and control of his garden, he bridged the gap between digital intelligence and physical resource management.
The Cost of Autonomy
While the setup sounds like a futuristic utopia, the execution provided a humbling lesson in AI efficiency. The agent's attempt to optimize the garden led to a comical failure in cost-benefit analysis, producing a single lettuce that cost roughly €60 to grow. It turns out that while agents are great at processing data, they still struggle with the "common sense" of resource spending in the real world.
The New Control Plane
This experiment highlights a growing trend: the shift from AI as a chatbot to AI as an operator. As lakehouses become the governance layer for identity, cost, and security, we are moving toward a world where agents don't just suggest a watering schedule—they execute it. We are currently in the "expensive lettuce" phase of this evolution, where the tools are powerful, but the guardrails are still being built.
Sources
Media
