01 · Solve
Start with the problem.
- Operational bottlenecks
- Customer friction
- Repetitive work
- Outdated processes
- Disconnected tools
- Digital gaps
Builder · Problem Solver · Operator
Austin Greene · ATXGreene
I build better ways for businesses to work.
From websites and automation to operations, AI, and custom systems, I find what is holding a business back and build what moves it forward.
AUDE FIERIDare to become.
The operating model
Solve. Build. Advance.
Understand the real problem, build the right answer, and leave the business more capable than I found it.
02 · Build
Create the right answer.
- Websites
- Automations
- Internal tools
- Dashboards
- AI systems
- Workflows & prototypes
03 · Advance
Make it work in the real world.
- Implementation
- Adoption
- Measurement
- Iteration
- Operational strategy
- Continuous improvement
What I do
What can we improve?
You don't need to know whether you need AI, a website, automation, or a process redesign. Describe what's happening — we'll figure out the right tool.
AI & Automation
Automate repetitive work, connect tools, deploy useful AI, and build intelligent workflows that save time or create capability.
Websites & Digital Systems
Websites, landing pages, portals, dashboards, digital experiences, and practical tools built around the way the business actually works.
Systems & Operations
Workflow design, process improvement, reporting architecture, internal systems, and operational problem solving.
Strategy & Execution
Turn a vague idea, messy problem, or opportunity into a practical plan — then help build and implement it.
Not sure what category your problem fits into? Good. Start with the problem.
Tell me what's stuck →For owners & operators
You don't need an “AI strategy.” You need a business problem solved.
- Maybe your website is outdated.
- Maybe your team spends hours moving information between systems.
- Maybe leads fall through the cracks.
- Maybe you've heard everything about AI but have no idea where it actually belongs in your business.
- Maybe you need a dashboard, internal tool, customer portal, workflow redesign — or something that doesn't have a name yet.
Start with the problem. We'll figure out the tool.
Selected work
Proof, not buzzwords.
Each project follows the same shape: the problem, what I built, and what it makes possible. The rest of the workshop lives in the Lab.
Operations & enablement
tools that make AI usable by real teams
snakepit.dev
- Problem
- Teams running AI agents watch costs and behavior drift with little visibility or control.
- Built
- An AI-operations command center: token-burn control, model routing, guardrails, and governance.
- Capability
- One place for a small or mid-market team to see spend, route models, and set guardrails before the invoice teaches the lesson.
Hermes Play Sheet
- Problem
- A capable agent stack is useless if the team doesn't know how to drive it.
- Built
- A single-page command and skill reference for agent operators.
- Capability
- New operators find the right command fast — enablement that turns a deployed tool into one people actually use.
AI systems
agent infrastructure, running as code
Mnemosyne
- Problem
- AI agents lose useful context between sessions and keep starting over.
- Built
- A persistent six-tier memory system with consolidation, retrieval, evaluation, and 19 model backends behind one API — pure-stdlib Python, 290+ tests.
- Capability
- Agents retain, retrieve, and compound what they've learned instead of resetting every time.
Tugboat
- Problem
- When an agent routes between models, tools, memory, and subagents, its decisions are hard to inspect or test.
- Built
- A declarative routing layer: each routing choice is one testable decision, with regret() as a reviewable “PR diff for routing.” ~1.2k lines, 34 tests, zero dependencies.
- Capability
- Agent behavior can be reviewed and tested before it ships — cleaner orchestration, safer iteration.
Range
interactive and immersive builds
Greene Halls
- Problem
- A dense body of research is hard to explore as documents alone.
- Built
- Twin walk-through WebGL exhibits in Three.js — 3D models, engraving backdrops, and narrated voice-over for 36 figures.
- Capability
- Research you can walk through in a browser — and proof of end-to-end range across 3D, audio, and interaction design.
Ways to work together
How we can work together
Start with one problem and stay focused — or widen the scope when the opportunity is bigger.
-
01 · Diagnose
Business Advancement Review
Identify bottlenecks, wasted effort, customer friction, digital gaps, and practical opportunities for technology or process improvement.
AI-specific? The AI Ops Diagnostic maps workflows, cost burn, and governance gaps into a 90-day adoption plan.
-
02 · Build
Focused Project
A website, automation, dashboard, workflow, AI system, internal tool, prototype, or another clearly scoped build.
-
03 · Transform
Systems & Operations Engagement
Modernize multiple connected workflows, tools, reporting systems, or operational processes.
-
04 · Advise
Ongoing Operator Advisory
Ongoing technical and operational problem solving for owners or teams that need a capable builder available as new problems arise.
About Austin
Operator brain.
Builder hands.
Researcher soul.
My work sits at the intersection of business, technology, operations, and building. I've spent years working directly with organizations and decision-makers while independently building AI infrastructure, software, automation, digital products, and experimental systems.
I don't believe every problem needs AI. I believe every problem deserves the right tool.
- 7+ yearsEnterprise SaaS, client operations, adoption, and executive-facing work
- Business + technologyB.B.A., Baylor University — Professional Selling · Entrepreneurship · International Business
- BuilderAI systems · automation · software · web · interactive experiences
- OperatorStrategy through implementation
Austin-rooted · remote-capable. Commercial engagements are delivered through Bluebonnet Legacy.
The ATXGreene Lab
Things I build because I want to know whether they can exist.
Deep technical work, research, games, and experiments. None of it is required reading to work with me — it's the proof that hard problems get figured out.
Agents & memory
One stack: Mnemosyne remembers, Tugboat routes, snakepit governs.
Mnemosyne
Six-tier agent memory with decay, offline consolidation, a human-approved self-improvement loop, and evals.
Tugboat
Declarative, testable routing for agent harnesses.
BEN
Agent navigator — a persona-forward front end to the Mnemosyne / Tugboat stack.
notes · coming soon
Eternal Context
Hybrid agent with persistent memory — the foundation Mnemosyne grew from.
Worlds & games
Everything generated or engineered in code, shipped to the browser.
Greene Halls
Twin walk-through WebGL exhibits with 3D models and narrated audio.
Shadow of the Watchers
Asymmetric browser RTS: fog of war, A* pathfinding, economy and tech trees, and a reactive AI opponent. Every model, sound, and map is procedural.
The Watchers: Fall of the Giants
Isometric action-roguelike built in Godot, with procedurally synthesized sprites, VFX, and music.
Training tools & simulation
Offline, zero-dependency progressive web apps.
Devil's Ace
Blackjack strategy trainer: drills with instant feedback, six counting systems, index deviations, and a Monte-Carlo edge simulator.
Above Black Trainer
Structured audio training: multi-band tone generation, progressive session protocols, and a session analytics dashboard.
Quantitative experiments & research
Markets, math, and the papers behind the systems.
APEX
Autonomous prediction-market research: cross-platform arbitrage and model validation on Kalshi and Polymarket, run paper-first.
private build · case study soon
Papers & preprints
Mnemosyne and Tugboat preprints, and a closed-form heat-kernel coefficient for transmission boundary conditions.
Latest from GitHub
Recently pushed public repositories, pulled live.
Writing & research
Thinking in public.
Essays, papers, and policy work on agents, memory, operations, and turning research into something usable.
-
2026
read ↗
Launchpad & the Civil Mobility Compact
A voluntary, rights-protected public–private model for rebuilding economic mobility in the AI era. Policy working draft.
-
2026
read ↗
I Built a Brain for My AI Agent. Here's What I Learned.
The Mnemosyne architecture in plain language — why memory must flow both directions, and how the dream cycle closes the loop. 14 min.
-
2026
preprint · soon
Mnemosyne: A Six-Tier Bidirectional Cognitive Memory Architecture
The paper — ICMS tiers, ACT-R decay, Hebbian reinforcement, and the Reflection→Instinct distillation loop. Zenodo DOI pending.
-
2026
preprint · soon
Tugboat: Declarative Four-Axis Routing for Agent Harnesses
Routing as a first-class, testable object — with regret() as a "PR diff for routing." Zenodo DOI pending.
-
2026
draft · soon
A closed-form a₃ for transmission boundary conditions
Heat-kernel coefficient for the pure-κ-jump transmission Laplacian. Targeting Letters in Mathematical Physics.
-
ongoing
read ↗
Substack
Working notes on agents, memory, trading, and applied math.
Contact
Tell me what's stuck.
You don't need a scope of work. Tell me what's happening, what you wish were happening instead, and we'll start there.