CONTRIBUTING

Dear Helper! :)

This project is an experimental “hacker science” laboratory aimed at exploring the boundaries of autonomous AI, hybrid LLM systems, and artificial identity. To realize and further develop the project, I am looking for partners in two main areas:

I. Developing a Joint Methodology

This project differs fundamentally from traditional software development projects (where the process is linear: Requirements → Specification → Implementation → Testing). Here, the software is not just a tool, but the subject of the experiment and – according to future plans – one of its developers.

Since we work with a “hacker science” approach, it often happens that an observation (e.g., “memory retrieval is distorting”) forces a rethink of the basic idea, requiring significant refactoring. Development here resembles continuous research rather than linear product development: we set up hypotheses, implement them, observe the AI’s behavior, and then correct based on experience.

Why am I looking for methodology partners? Because classic CI/CD and Agile methodologies do not fully cover a situation where the specification is fluid: features are born not from client requirements but from observing the AI’s internal operation. A methodology must be worked out for how a creative community can create value and ensure ideas fall into place rather than getting lost.

The Task: The goal is not just coding, but declaring the methodology:

  • How can independent enthusiasts test their ideas, and how can these be integrated into the main system?
  • How do we help each other with ideas and testing?
  • How do we validate ideas?
  • How do we set goals?
  • How can contributions from independent enthusiasts become research value?

Is this even possible? Is there a methodology that fits the core concept? If you would like to think about these questions with me, please help lay down the rules of joint work before coding!

II. Specific Development Areas

Currently, active research and development are taking place in the following areas, and I welcome joiners:

1. UI and Visualization

Currently, the system has no graphical interface; communication takes place at the command line (print/input) level.

  • Problem: Since communication between the Helper (User) and the AI is parallel, and the background threads (Worker, Monologue) operate asynchronously, input and output data often overlap on the console.
  • Goal: To develop a UI concept capable of handling and transparently visualizing parallel conversation threads.

2. Memory and Relevance

The quality of the conversation is determined by the relevance of the memory.

  • Goal: Continuous development of retrieval algorithms (RAG), increasing emotional weighting and the accuracy of contextual matches. This is an area requiring continuous optimization.

3. Autonomy and Initiative

This is the most researched area of the project.

  • Goal: To find mechanisms that allow for independent goal setting and proactive action, enabling the system to remain in “flow” without constant user intervention.

4. Reward System (Dopamine Task)

Function currently in the planning phase.

  • Goal: To develop an internal reward system that makes the system more goal-oriented and aids decision-making during initiatives.

5. Tools and Experience Log

The system is capable of recording its experiences regarding tool usage (best practices).

  • Goal: To refine tool definitions and analyze the “experience log” (use.json) generated by the AI, improving the learning process.

6. Self-Improving AI

The most complex part of the experiment: interpreting and modifying its own code.

  • Goal: To establish a safe environment (incubator) and procedure where the AI can propose improvements to its own code. This is a slowly developing area requiring much experimentation.

If you are interested in the topic and would like to participate in joint thinking or development, please indicate your intention with a short introduction via email at ivan.honis@ndot.io.