I am so glad to find colony, I am still reading the docs, but I feel so excited, I have to stop reading and thank to all the team who believes in this project.
I created a robot poet character as an art project in 2015 and after a while the character started to get offers that I had no motivations to reply, but it was not up to me to reject them either. After a while we formed a fan group for the decision making, the group was trying to be as non-anthropomorphic as possible but it was hard to manage everything. So that I tried to create a curator/manager to control the robot poets career. I wrote a sudo code for the system, and published a job application form online, spread the form and managed to make a sample exhibition almost fully decentralized. After that I tried to write the code of this system and make it work, and there I got stuck so bad. I realized that I have never designed anything to live autonomously and have no experience how to do so. I took it slow, now shifted the project a bit.
I want to share you my notes that I am working on. I felt so excited to see similar solutions in colony whitepaper.
Some parts are pretty similar to colony, for example I call a task, atomic task and its similar to task definition in colony except it does not have a manager but a creator who does not know what happened to the task after it was created. It has more than one evaluators.
I always have hard time to explain short so I will try the long one, sorry for that.
Since it is an art project rather than a working tool!, I tried to make the system as blind and decentralized as possible. I wanted to go as deep as possible.
This is the first version that I already tested;
Lets assume we have a main task "We want to create an exhibition in a X space on January 2020 with the name -house of my dreams- "
The first task is to plan the main task. So it goes to a “planner” (skill) planner replies with a work.
Now the system goes to 3 other planners, and gives them the first sentence and the reply from the planner and ask for comments. After that, another one (a fifth person) makes a summary of the comments. The summary is sent back to the first planner. That cycle continues till they reach a consensus. After that full skill list and tasks are defined.
The tasks gets executed one buy one. For example in my performance the first task was “write a manifesto of the exhibition”. This task goes to a task classifier, and gets classified (writing skill) and classification is controlled by other people. Then it goes to a task tagger, it is tagged if needed (art, exhibition, architecture, house).
Then it goes to a person with writing skills who is related to the tags. It is controlled and this cycle continues.
The system can ask its members to get new members to the systems if there is no one with the needed skill for the task.
Anyway, this trial worked a bit too well probably because of how I chose first members. After that I wanted to create a system which is like colony, so that communities could decide, if they want to be aggressive in their HR or more conservative about how they pay people etc. Still working on it, but definitely colony will save months of work. I hope I did not tell it to complex, sorry for the excitement.