Skip to content

Hasatel/PSLRAR

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 

Repository files navigation

PSLRAR — Prototype of Self-Learning Recursive Artificial Retardation

A recursive self-improving code-based memory and generation system.

Status

Prototype — early concept stage. Expect instability, experimentation, and fun bugs.


Core Idea

PSLRAR is an experimental project aiming to simulate a form of recursive memory, data association, and code evolution. It works not only with knowledge of code fragments, but with associative memory and meta-level mutation of its own logic.

The system is designed to:

  • Store code fragments as meaningful, context-rich objects.
  • Connect fragments with associative keys ("smell of rain" model).
  • Retrieve, combine, mutate, and improve code pieces.
  • Remember what worked — and what didn’t.
  • Evolve its own behavior and logic over time.

This is not a traditional AI/ML system. It's a sandbox to explore a weird idea: what if memory, pattern-matching, and code-gen lived together — and kept rewriting each other?


Core Components

Code Fragment

Basic unit of information. Each fragment contains:

  • The actual code.
  • Metadata (language, purpose, dependencies).
  • Importance weight.
  • Contextual keys (e.g. "file reading", "error-prone", "fast loop").
  • Links to other fragments (influences, usage history, etc).

Associative Keys

Like smells or visual memory. They are fuzzy, descriptive handles like:

  • "recursion", "I/O", "unsafe", "clean but slow", "async"
  • Used to connect, retrieve, and mutate fragments.

Self-Modifying Engine

The engine does:

  • Retrieves fragments by context.
  • Merges, rewrites, replaces code.
  • Generates new code when needed.
  • Evaluates result (via tests or heuristics).
  • Updates memory based on success/failure.
  • Can rewrite its own logic.

Recursion and Self-Evolution

The project is recursive at its core:

  • The memory system can modify its own structure.
  • The generation engine can rewrite its own generation methods.
  • Old data can decay or be overwritten.

Planned Features

  • Graph-based fragment storage.
  • Key weighting and fuzzy matching.
  • Mutation and variation engine.
  • Simple compiler/tester loop.
  • Feedback-based memory shaping.
  • Visualization of memory clusters.

Disclaimer

This project is experimental, absurd, and potentially useless — but fun. It is not based on any formal AI architecture. It is a prototype of something in between madness and code art.


Repo Structure (planned)

  • src/ – code for memory engine and logic
  • fragments/ – initial hand-written fragments
  • tests/ – simple scenarios to test generation
  • docs/ – design ideas, visualizations, rambling

License

TBD. Possibly MIT, possibly something weird.


Author

Initiated by @Hasatel — inspired by strange dreams, recursive thoughts, and my schizophrenia.


“The system forgets nothing. But it forgets how it remembered.”

About

Prototype of Self-Learning Recursive Artificial Retardation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published