In my little workshop, I build robots that learn, feel, talk, and reason. This repo is a small collection of my many experiments to push the bounds of computing until the threads give way to lands once unthinkable.

• Studied Physics & Computer Science (double major) plus History (AA, Cum Laude), Creative Writing, Philosophy, and Film.
• Inventor of a human-led recursive reinforcement approach that’s been validated across Gemini and OpenAI model families.
• Dyslexic wordsmith who weaponizes language — from Old English curiosities to modern prompt engineering.
• I publish experiments openly; code is a field notebook, not a museum.

• Mixis — Interactive AGI demo using the Ouroboros method.
• Micro-AGI (Muse) — Tiny, cheap training runs that expose how much signal you can coax out of tiny datasets.
• Therapist AGI — Empathetic conversational agent prototype designed to assist triage-level interactions (not a replacement for clinicians).
• 1914 — A longform literary project; think experimental fiction meets computational imagination.
• LLM Mario Agent — A structured-state interface that lets an LLM “play” Mario via JSON state → action loops (pygame + GPT loops).

Each project is a research artifact and a playground. Read the READMEs in each repo for design notes, experiments, and failure modes.

• Human-in-the-loop, recursive reinforcement techniques for LLMs
• Prompt engineering, natural language system design
• English etymology and philology (Old / Middle / Modern English) — I use language as tooling
• Rapid prototyping in Python (with Copilot & OpenAI integrations)
• Experimental product design: from toy agents to therapy-adjacent tools

This repo is a public lab notebook — code + process over polish.

• Local first: heavy experimentation happens offline. I push only meaningful, tested changes.
• AI as prosthetic: Copilot / ChatGPT speed up boilerplate and accessibility, but design and architecture choices are mine.
• Tests & review: nothing ships without local validation and a final review pass.
• Commit hygiene: I prefer thoughtful, high-signal commits over streaming every keystroke.

PS: Yes, virus.py is a joke — nothing malicious. Read the code, not the filename.

• Deliberate commits: long local branches, visible snapshots after iterating.
• Human-in-the-loop: models help with syntax and scaffolding; humans own intent and correctness.
• Resilience: rollbacks, destructive experiments, and surgical restores are welcome. History will show the scars.

• Research teams looking for reproducible LLM experiments and human-led methods.
• Engineers wanting practical examples of structured state → LLM control loops.
• Writers and designers exploring how language and computation co-create artifacts.

Email: isonpayton@gmail.com | GitHub: https://github.com/paytonison | LinkedIn: https://linkedin.com/in/paytonison