Every now and then, when I sit down to write one of these recaps, my first thought is: “This will probably be a short one this week. Not much has happened.” In almost every case, I still end up having a few noteworthy experiences to share. When you travel around to new places, you tend to see something new.

This week, I haven’t travelled around. I haven’t explored. I haven’t really been out doing things either. This week, I have buried myself in a project on my computer and built a fairly structured routine around it.

My day starts in a small private room in a bungalow on Koh Pha Ngan. I have a nice soft bed, clean towels, my own bathroom, a fridge, a small desk, a hammock and, generally, some peace and quiet. I booked the bungalow for 14 days and will be staying there until August 30. My morning consists of a bit of stretching, a few push-ups, a few squats and then I’m off. I pack my computer and leave the bungalow.

On my way towards Thong Sala, I find somewhere to have breakfast. There are several good cafés, but I have found a breakfast café run by a Portuguese man. He bakes his own bread, which tastes fantastic, and he doesn’t charge much for breakfast. His cats come over to say good morning. They are incredibly fond of the guests, and if you give them any attention, you’ll need help getting rid of them when you eventually want to leave.

I arrive at my work café. The staff know my order. A large Americano. Coffee in hand, I sit down at a desk. It’s an IKEA desk with an IKEA office chair, but that is still a much better ergonomic setup than you’ll find in most cafés suitable for working. I sit there until they close in the afternoon. Some days I leave a little earlier. Other days I stay right up until closing time.

Then I walk back to the bungalow. I put on some entertainment and keep working for a while until dinner. For dinner, I head out to one of the restaurants in the area. The bungalow hotel does have its own restaurant. It does the job if I’m tired or really just want to get back to my computer, but otherwise I go out and find a restaurant I haven’t tried yet.

After dinner, I head back to the bungalow, where the plan is simply to relax for a bit and watch some videos. Usually, I end up working a little more while watching them.

Around midnight, I decide that I should probably go to bed. Shower, teeth, toilet and then into bed. That has pretty much been my life for the past week and a half.

This week’s project

At this point, someone is probably wondering: “What the hell are you working on?” It feels slightly embarrassing to write about because it is incredibly nerdy, incredibly strange and a little controversial.

I’m building a self-sustaining AI setup. It costs money to use the good AI models, but what if we could make them pay for themselves? It needs to be able to generate its own income with little to no involvement from me. It needs to be able to assess what things cost and handle paying for its models, infrastructure, expenses and general upkeep.

I have taken what I know about software development, product development, organisations and businesses and boiled it down into a set of rules that allow AIs to plan, design, execute and maintain themselves over a long time horizon. Much like an organisation. It isn’t one AI functioning as an organisation, but multiple AIs working within an organisational structure. If the project reaches the point where it can pay for itself, it will become interesting to see how far we can push the organisation’s authority and autonomy.

There is a lot of work involved in setting up the project, but once the basic structure is in place, it becomes a matter of managing the AI models, as they will be the ones building it further from there.

It might sound a little mysterious and a little sci-fi, but based on everything I have seen and experimented with over the past six months, this should be more than possible with what is available today. The biggest challenges are making sure that rules and laws are followed, because I can’t just point at the AI box and blame it. Ultimately, I am responsible for its actions. So a large part of the work is actually about building something safe enough that I’m willing to let it loose.

My hope is to have the first version ready before I leave Koh Pha Ngan, but the project has grown considerably and has already swallowed many more hours than I initially thought it would, so we’ll see.

From here

It is fairly typical for me to find something and become hyperfixated on it for a while. This project is particularly stimulating because it combines many of my interests into one thing. In a way, it builds on some of my earlier thoughts about temporary bases. That is also why I have deliberately allowed myself to try out a proper working week here on Koh Pha Ngan.

Whatever happens with the project, I’m leaving Koh Pha Ngan on August 30 and heading to Koh Tao. I still plan to get my diving certification on Koh Tao before moving on.

I said goodbye to Andy earlier this week. He continued on to Koh Tao. By the time I get there myself, he will probably already have moved on. It does look like we might be in Chiang Mai at the same time, though. We have at least agreed to stay in touch and see if we can meet up again somewhere further down the road.

But I’m not going to make those plans yet. My approach has always been to keep my plans as loose as possible. That said, I’m starting to have a general idea of how the final months of my trip will play out. At the end of September, I’m heading back to Bali. From Bali, I’ll travel east and spend most of October in that part of Indonesia. In November, I’ll head to the Philippines, and from there I’ll return to Denmark at the beginning of December.

For now, I’m diving back into my little AI project.

Want to support the journey? Read more here.