Skip to content
View leerobber's full-sized avatar
🎯
Focusing
🎯
Focusing
  • DeNyse
  • Douglasville

Block or report leerobber

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 is supported. This note will only be visible to you.
Report abuse

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

Report abuse
leerobber/README.md
████████╗███████╗██████╗ ██████╗ ██╗   ██╗    ██╗     ███████╗███████╗
╚══██╔══╝██╔════╝██╔══██╗██╔══██╗╚██╗ ██╔╝    ██║     ██╔════╝██╔════╝
   ██║   █████╗  ██████╔╝██████╔╝ ╚████╔╝     ██║     █████╗  █████╗  
   ██║   ██╔══╝  ██╔══██╗██╔══██╗  ╚██╔╝      ██║     ██╔══╝  ██╔══╝  
   ██║   ███████╗██║  ██║██║  ██║   ██║        ███████╗███████╗███████╗
   ╚═╝   ╚══════╝╚═╝  ╚═╝╚═╝  ╚═╝   ╚═╝        ╚══════╝╚══════╝╚══════╝

Terry Lee — Systems Architect · Autonomous AI · Sovereign Infrastructure

Self-taught. Self-funded. Self-improving — just like the systems I build.


Sovereign Core KAIROS Hardware DGM


What I Build

I build autonomous AI systems that improve themselves — sovereign infrastructure that runs entirely on local hardware, evolves nightly, and answers to no cloud provider, no institution, and no gatekeeping API.

No team. No funding. No shortcuts.

Just architecture, compounding work, and a home GPU cluster that never sleeps.

"The infrastructure doesn't care about credentials. Neither do I."


The Sovereign Stack

This isn't a collection of projects. It's one unified system.

┌─────────────────────────────────────────────────────────────────┐
│                     SOVEREIGN CORE                              │
│              Heterogeneous Compute Gateway :8000                │
│         Routes intelligence across 3 processors                 │
├──────────────────┬──────────────────┬───────────────────────────┤
│   RTX 5050       │   Radeon 780M    │      Ryzen 7 CPU          │
│   8GB VRAM       │   4GB VRAM       │      Fallback             │
│   Qwen2.5        │   DeepSeek-Coder │      Llama / Mistral      │
│   :8001          │   :8002          │      :8003                │
└──────────────────┴──────────────────┴───────────────────────────┘
           │                │                    │
           └────────────────┼────────────────────┘
                            │
              ┌─────────────▼─────────────┐
              │      KAIROS / SAGE        │
              │  Self-Improvement Engine  │
              │  10 live cycles / night   │
              │  Proposer→Critic→         │
              │  Verifier→Meta-Agent      │
              └─────────────┬─────────────┘
                            │
        ┌───────────────────┼───────────────────┐
        │                   │                   │
   ┌────▼────┐        ┌─────▼─────┐      ┌─────▼──────┐
   │   DGM   │        │HyperAgents│      │contentai   │
   │ Darwin  │        │Self-refer.│      │    -pro    │
   │ Gödel   │        │  Swarm    │      │8-platform  │
   │ Machine │        │           │      │  engine    │
   └─────────┘        └───────────┘      └────────────┘
        │                   │                   │
        └───────────────────┼───────────────────┘
                            │
              ┌─────────────▼─────────────┐
              │          Honcho           │
              │   Mission Control UI      │
              │  Live GPU · KAIROS Viz    │
              │  Swarm Nexus · LatentLog  │
              └───────────────────────────┘

Every repo is a node. The system is the point.


Core Projects

Project What It Does Stack
sovereign-core Central hub — GPU gateway, KAIROS SAGE loop, WebSocket event bus, metrics Python · FastAPI
DGM Darwin Gödel Machine — AI agent that evolves its own source code Python
HyperAgents Self-referential agents that rewrite themselves to optimize for any task Python
Honcho Mission control dashboard — live GPU health, KAIROS score viz, voice interface TypeScript · React
contentai-pro Multi-agent content engine with adversarial debate, DNA fingerprinting, 8-platform output Python · FastAPI · Docker
Termux-Intelligent-Assistant AI-embedded Android terminal agent — works fully offline Python

How It Works At Night

While I sleep, the system doesn't.

03:00 AM ──► KAIROS fires
             10 live ARSO cycles on RTX 5050
             Proposer generates improvement proposals
             Critic reviews · Verifier checks logic
             Best proposals archived as stepping stones
             Every 3 cycles: Meta-Agent rewrites the rules

04:00 AM ──► 13-Amplifier Training Engine
             Memory expansion · Adversarial drills
             Concept synthesis · Goal calibration
             Sub-agent stress tests · Ghost Protocol training
             Cross-domain transfer · Identity verification
             Before/after logged to persistent memory

             Total: ~3,650 live improvement cycles per year
             Compounding nightly. No days off.

The Philosophy

Most AI systems are static. You use them the same way in year one as in year five.

I build systems that compound.

Inspired by two ideas:

  • Darwin — keep good solutions, build on them, improve generation by generation
  • Gödel — a system powerful enough can reason about itself and rewrite itself to be better

Put those together and you get infrastructure that evolves itself, using itself, every single night.

I'm not trying to build the smartest system today.
I'm building the one that gets smarter fastest over time.


Technical Fingerprint

terry_lee = {
    "location":     "Douglasville, GA",
    "approach":     "self-taught, systems-first, local-first",
    "languages":    ["Python", "TypeScript", "GhostScript*"],  # *custom-built
    "hardware":     ["RTX 5050 (8GB)", "Radeon 780M (4GB)", "Ryzen 7 CPU"],
    "philosophy":   "sovereign infrastructure — no cloud dependency, no gatekeepers",
    "building":     "autonomous AI that improves itself while I sleep",
    "status":       "compounding daily",
}

Self-taught. Self-funded. Self-improving — just like the systems I build.


What I'm Working Toward

  • Gateway live on TatorTot — first real KAIROS inference cycles
  • DGM stepping stones feeding the KAIROS archive
  • Honcho dashboard fully connected to live system
  • GhostScript async/await — concurrent agent execution
  • GH05T3 designing her own training agenda autonomously
  • Sovereign stack fully wired across 5 repos
  • KAIROS SAGE loop — 4-agent co-evolution implemented
  • Heterogeneous compute gateway — 3-processor routing
  • Custom AI-native programming language (GhostScript)
  • Nightly 13-amplifier training engine running

The work exists. Now you can see it.

leerobber · Douglasville, GA · Building in the dark since before it was obvious

Pinned Loading

  1. VAGEN VAGEN Public

    Forked from mll-lab-nu/VAGEN

    NeurIPS 2025 — Training VLM agents with multi-turn reinforcement learning. Tracking for integration with HyperAgents and sovereign-core vision systems.

    Python