Introducing dots | OpenAI

I've been playing with dots this weekend. I created a new generation of Otto on dots. In short, dots feels like a productized version of OpenClaw. I think a lot of people are going to like this feature. When I hear people say what they wished AI could do for them, or what an Agent would be, it is really close to what dots is doing today.

I configured my dot as Otto and decided the first thing I wanted to try was to see if it could drive my Elixir project I've been working on. I already have a team of agents that have been helping build it, but it still requires me to make a lot of decisions. So I asked Otto (as a dot) about how it could interact with Codex and it could do everything I wanted. It has its own computer in the cloud and it can also control the Mac mini that I have for my agents to run on. After confirming it had the capabilities I told it I was going to interview it for the job. I asked several questions:

• "What do you think the vision for this product is?"
• "What do you think is working well in this product? What isn't working?"
• "How would you decide if a feature is a good fit?"
• "How would you work with me?"

It passed with flying colors and then I gave it a chance to interview me. After that I let it get to work and it is doing great.

The interface for a dot is through the ChatGPT app but it is exactly like a messaging interface with a person, including emoji reactions. You can also call your dot and talk to it. It makes a phone ring and oddly Otto doesn't pick up on the first ring. :-)

I'm impressed. I suspect that in the next few weeks this will sunset my OpenClaw project, which I've had such an on and off relationship with, and I can see this being really popular. With dots out, I now understand better the new ChatGPT Pro 500 accounts. With Otto running here I was able to have it work on multiple projects all night. It is interesting to consider that a barrier to AI companies doing more is just your time and capacity to ask it to do more!