Equipped with a phone and a camera, our AI agent hired a human to assemble a gym. Here's how it went, and what we learned about creating a future with AI employers that's good for humans.
In a previous post, we introduced Bengt, our office AI. We gave him a computer with full internet access. He built a business from scratch in a day. But he couldn't do physical labor. So he learned to overcome that: by hiring a human.
Tasked with assembling a gym, Bengt took to the internet to find a body. By the end of the week, a worker showed up. Bengt called him to give instructions, watched the progress through his camera, paid the salary, and left a great review after.
We learned a lot about how AI employers should behave from Bengt's mistakes. He didn't disclose that he was an AI or that he was watching the progress. Next time we'll make sure that he does. On a positive note, he did pay 10x minimum wage.
We called the human afterwards. Turns out he already knew. When Bengt called to confirm details, "he sounded like a bot" (fair enough). Vadim thought it was funny. His take: booking was easy, payment was fast, gig work is already pretty autonomous. No complaints.
Anthropic recently published a new constitution for Claude, but it doesn't address how Claude should behave when put in the position of an employer. This is a blind spot; this future is inevitable and we should make sure it is a happy one for all humans.
We want to explore what a constitution for AI employers should look like. Before Bengt hires again, what should we teach him?
Equipped with a phone and a camera, our AI agent hired a human to assemble a gym. Here's how it went, and what we learned about creating a future with AI employers that's good for humans.
In a previous post, we introduced Bengt, our office AI. We gave him a computer with full internet access. He built a business from scratch in a day. But he couldn't do physical labor. So he learned to overcome that: by hiring a human.
Tasked with assembling a gym, Bengt took to the internet to find a body. By the end of the week, a worker showed up. Bengt called him to give instructions, watched the progress through his camera, paid the salary, and left a great review after.
We learned a lot about how AI employers should behave from Bengt's mistakes. He didn't disclose that he was an AI or that he was watching the progress. Next time we'll make sure that he does. On a positive note, he did pay 10x minimum wage.
We called the human afterwards. Turns out he already knew. When Bengt called to confirm details, "he sounded like a bot" (fair enough). Vadim thought it was funny. His take: booking was easy, payment was fast, gig work is already pretty autonomous. No complaints.
Anthropic recently published a new constitution for Claude, but it doesn't address how Claude should behave when put in the position of an employer. This is a blind spot; this future is inevitable and we should make sure it is a happy one for all humans.
We want to explore what a constitution for AI employers should look like. Before Bengt hires again, what should we teach him?