Member-only story
No human was allowed to intervene — just an AI, a webcam, and a real emergency. The viral version of this story gets the facts wrong; the true one is better. #
Somewhere right now, a tomato plant named Sol is growing under a grow light — and no human is taking care of it.
Its caretaker is Claude — Anthropic’s AI model, though notably not in an Anthropic experiment — which wakes up on a schedule, reads the sensors, looks at the plant through a camera, decides whether Sol needs light, heat, air, or water, and acts. There’s a live dashboard where anyone can watch: the webcam feed, the sensor readings, and — this is the part I love — Claude’s actual reasoning, logged decision by decision. As I write this, the dashboard reads Day 38.
You may have seen a version of this story going around — “AI saves dying plant in 13 minutes!” — and like most viral AI stories, the retellings have mutated the facts. So let me tell it properly, with what’s actually verifiable, because the true version doesn’t need embellishment.
The experiment #
First, a correction to the most common error in the viral versions: this is not an Anthropic…