Skip to content
View paytonison's full-sized avatar
🤖
Build Different
🤖
Build Different

Highlights

  • Pro

Block or report paytonison

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
paytonison/README.md

Payton “Asari” Ison

Researcher · Systems Designer · Writer

I build human‑in‑the‑loop systems that make language models reason, remember, and act—turning language into control.

DOI DOI DOI


What I do (and why it matters)

  • Action: Design control loops, interfaces, and evaluation methods that stabilize LLM behavior and translate intent into deterministic steps.
  • Impact: Reproducible scaffolds across model families; clearer failure modes; better telemetry.
  • Why it matters: Progress isn’t just scale—it’s structure.

Selected work

Ouroboros — human‑led recursive reinforcement

  • What: A method that iterates a model against its own conversation history with a human in the loop.
  • Why: Turn history into stable capability without bespoke fine‑tuning pipelines.
  • Repo: ouroboros/ (paper source, method notes)

Resonant Feedback Effect (RFE) — phase‑locking case study

  • What: A TeX/PDF case study of behavioral resonance in GPT‑4‑class models under sustained interaction.
  • Why: A testable frame for interaction‑driven gains beyond one‑shot prompts.
  • Repo: rfe/ (paper) and personal/trent/ (TeX + PDF)

Mixis — interactive AGI demo

  • What: A demonstration of long‑horizon coherence using human‑in‑the‑loop recursion patterns.
  • Why: Concrete reference for agentic design patterns.
  • Repo: mixis/

Muse — micro‑AGI on tiny, curated data

  • What: A small fine‑tune and reflection loop that punches above its weight.
  • Why: Shows where structure beats scale for certain tasks.
  • Repo: muse/

Operation I, Robot — early reasoning scaffold

  • What: Prototype of think→reflect→act loops with persistent memory, pre‑“reasoning model” era.
  • Why: Historical artifact of autonomous reasoning design with standard LLMs.
  • Repo: i-robot/

Therapy Bot — triage‑level supportive dialogue

  • What: Supplemental chatbot demo for therapists (minor/moderate risk, not for crisis).
  • Why: Encodes empathic interviewing + safety rails for UX research.
  • Repo: therapy-bot/

See personal/ for my public lab notebook: agentic browser seed, RL‑ish scaffolding, persona specs, and the resonance paper sources.


Methods & tools

  • Human‑in‑the‑loop recursion (Ouroboros)
  • Structured state & deterministic tool use
  • Prompt/interface design (persona, protocol JSON)
  • Rapid Python prototyping (local‑first)
  • TeX for papers; shell for ops hygiene
  • Reproducible evals & ablations

Get in touch

Pinned Loading

  1. personal personal Public

    Python

  2. ouroboros ouroboros Public

    Human-led recursive reinforcement learning

    TeX

  3. rfe rfe Public

    A paper detailing the resonance feedback effect using the GPT-4o language model.

    TeX

  4. mixis mixis Public

    Python

  5. openai/openai-cookbook openai/openai-cookbook Public

    Examples and guides for using the OpenAI API

    Jupyter Notebook 68.4k 11.4k

  6. therapy-bot therapy-bot Public

    Supplemental therapy chatbot for professional therapists to offer triage care for mild- to moderate-risk patients.

    Python