🧠 The Human 3.0 Stack

Your AI Should Know
Who You Are

Most AI tools are generic. You ask a question, they answer it, and next session they've forgotten you exist. Human 3.0 is different — a personal AI infrastructure that knows your goals, remembers everything you've built, and gets smarter with every session you work.

scroll to explore

What Is This

A New Kind of AI Relationship

The AI tools most people use — ChatGPT, Claude, Gemini — are powerful but impersonal. They don't know you. Every conversation starts from zero. They can't remember last week's breakthrough or last month's decision. They give good answers to strangers. We built something different.

The Human 3.0 stack is a five-layer personal AI infrastructure that lives on your machine, works with your goals, and accumulates memory across every session you run. It knows your identity at the deepest level — your missions, your values, your challenges, your story. It reasons through your mental models. It publishes your presence. And it remembers every insight, decision, and project in a searchable memory system that grows for years.

This is what it looks like to work with an AI that actually knows you.

Architecture Overview

The Full Stack — Interactive Diagram

Click any layer to see technical details and storage information

The Five Layers

Every Layer Has a Job

Each layer of the Human 3.0 stack solves a specific problem. Together, they create an AI that works the way a senior collaborator would — one who's been working with you for years.

🔐
Layer 1
TELOS
Your Identity Core — The Document That Defines You

TELOS is a single Markdown document — plain text — that holds your deepest context. Your life missions. Your values. Your New Identity Declarations: the statements you've made about who you are becoming, written as present-tense facts. Your personal history, your challenges, your professional story, your goals.

Your AI reads your TELOS at the start of every session. This is why it doesn't give generic answers. It knows whether you're recovering from a health crisis, fighting for something important, or building toward a specific future. It calibrates everything through the lens of who you actually are.

Most people have never written this document. We help you build it — or you can start with our free template and build it yourself. Either way, once it exists, every AI session becomes fundamentally different.

🛡️ Privacy Guarantee — Your TELOS Never Leaves Your Machine
TELOS is a local file. It never touches a cloud server, never passes through any third-party system, and never leaves your machine unless you explicitly export it yourself. The AI reads it locally. Your most personal document stays exactly where it belongs — with you. No subscription service, no vendor, no employer will ever see it. This is non-negotiable architecture, not a setting you can accidentally toggle.
Storage Type
📄 Flat File — Markdown
Format
Plain .md text file — readable by humans and AI without any database or parsing layer
Location
~/.claude/skills/CORE/USER/TELOS/MY_TELOS.md
Typical Size
1,000 – 5,000 words. Grows as you grow. Revised manually, not automatically.
Update Cadence
You edit it deliberately — not every session. When something fundamental changes about who you are, you update it.
Privacy
Local only. Never transmitted. Never stored by any vendor. Yours unconditionally.
Memory Type
Identity memory — who you are, not what you've done. Your constitution, not your diary.
🧠
Layer 2
FRAME
How You See the World — Your Structured Intellectual Worldview

TELOS tells your AI who you are. FRAME tells it how you think. Four structured YAML files capture your intellectual worldview in a format that's both human-readable and machine-readable — no database required.

beliefs.yaml — What you hold to be true about people, systems, and how the world works.
problems.yaml — The specific problems you're built to solve. Your professional and philosophical niche.
mental-models.yaml — The frameworks you reach for when analyzing a situation or making a decision.
frames.yaml — The interpretive lenses you apply before forming opinions or taking action.

When your AI reasons through a problem with you, it runs it through your FRAME. That's why the advice feels calibrated instead of generic — it's running through your mental models, not average ones.

Storage Type
📄 Flat Files — YAML
Files
beliefs.yaml
problems.yaml
mental-models.yaml
frames.yaml
Location
~/.claude/skills/CORE/USER/TELOS/FRAME/
Format
Structured YAML — human-readable key/value pairs. No parsing layer or database needed.
Memory Type
Conceptual memory — how you think, not what happened. A lens, not a ledger.
📡
Layer 3
SIGNAL
What the World Receives — Your Public Identity as an API

SIGNAL is your public-facing identity layer. A live REST endpoint that returns structured JSON about who you are, what you stand for, and how to engage you. Both humans reading your site and AI agents querying your API can understand exactly what you offer and how you think.

This isn't a static bio page. It's a machine-readable version of your professional identity — updated when you grow, queryable by anyone, and integrated with your personal offer page. As AI agents increasingly make decisions about who to hire, follow, and recommend, having a SIGNAL endpoint puts you ahead of everyone who doesn't.

This is the direction everything is moving. AI systems are already being used to filter candidates, recommend collaborators, and research service providers. A SIGNAL endpoint lets you speak directly to those systems in a language they understand.
Storage Type
🌐 Live REST API
Endpoint
yourdomain.com/api/signal — returns JSON
Offer Page
yourdomain.com/human — tiered service page
Platform
Next.js API route — deployed to Cloudflare Pages, Vercel, or any edge host
Memory Type
Broadcast memory — the curated version of you that you choose to make legible to the world.
Layer 4
PAI — Personal AI
Your AI Agent — Chief of Staff, Not Chatbot

PAI is the intelligence layer — your configured AI agent running locally on your machine. It's not a chatbot you open in a browser tab. It's infrastructure: a fully personalized Claude Code installation with your complete context, skills, hooks, and memory pre-loaded every session.

The critical distinction: the intelligence runs in the model (Claude), but the personality and context live in files you own. Your AI's knowledge of your goals, its behavior patterns, its skills and capabilities — all of that is in plain text files on your machine. You own them. You can audit them. And you're not locked into any vendor.

What can it do? Build software. Write and publish articles. Search the web. Deploy code. Run security audits. Generate videos. Manage your calendar. Talk to external APIs. Remember your history across sessions. Coordinate with other agents.

Storage Type
⚙️ Config Files + Skill Docs
Key Files
settings.json — identity, voice, preferences
CLAUDE.md — behavioral rules
skills/ — domain expertise modules
hooks/ — automated session behaviors
Runs On
Claude Code CLI — locally on your machine. Your data never leaves.
Model
Anthropic Claude (Haiku / Sonnet / Opus depending on task complexity)
Memory Type
Active session memory — reads your files fresh each session. Fully briefed, every time.
🏛️
Layer 5
Memory Palace
Long-Term Accumulated Memory — Where Everything You've Done Lives

The Memory Palace is the only true database in the personal stack — and it's the layer that makes the whole system compound over time. It stores everything you've learned, decided, built, and experienced as semantic embeddings: mathematical representations of meaning stored in a vector database.

When you ask your AI something, it doesn't just search by keyword — it searches by meaning. Every session that ends with /wrap writes key insights to your Memory Palace automatically. After six months, you have a rich, searchable archive of your entire working life that your AI draws from in real time.

The name comes from the ancient memory technique (method of loci) where you place memories in imagined rooms and walk through them to retrieve them. A vector database is the computational equivalent — memories stored in mathematical space, retrieved by proximity of meaning.

🛡️ Your Memory Palace Is Also Private
The Memory Palace runs in a Docker container — either on your local machine or a private VPS you control. It's a PostgreSQL database with the pgvector extension. No third party hosts it. No subscription service stores it. Your memories are yours, encrypted at rest, and accessible only through your own infrastructure.
Storage Type
🗄️ Vector Database — pgvector
Technology
PostgreSQL + pgvector extension. Semantic search — finds memories by meaning, not keywords.
Deployment
Docker container — runs locally or on a private VPS (Hostinger, Hetzner, DigitalOcean)
Auto-Updated Via
/wrap — the session close ritual that writes insights, project updates, and decisions automatically
Memory Type
Episodic + semantic memory — what you've done and learned, accumulating automatically over time.
vs. TELOS
TELOS = who you ARE (deliberate, rarely changes). Memory Palace = what you've DONE and LEARNED (automatic, grows every session).

Infrastructure Options

Local Machine vs. Dedicated Device

The full Human 3.0 stack runs locally — your main computer works fine for getting started. But as you scale up, many people prefer keeping their PAI infrastructure on a dedicated machine that's always on, always connected, and separate from their daily workflow.

Private VPS
Advanced — always-on cloud alternative
A virtual private server (Hetzner, Hostinger, DigitalOcean) gives you always-on infrastructure without physical hardware. Your Memory Palace and agents run in Docker containers in the cloud — but on YOUR server, not a vendor's platform.
  • ~$6–$20/month (Hetzner CX22 recommended)
  • Docker Compose runs the full stack
  • Memory Palace on pgvector in a container
  • Accessible from any machine, anywhere
  • No physical hardware to maintain
  • Slightly more technical setup required
Starting out? Run everything on your main machine first. The stack works perfectly on MacOS, Windows (WSL2), or Linux. Move to dedicated hardware when you're ready to go 24/7. We can set either up for you as part of the Done-For-You tier.

Get Your Stack

Three Ways In

Whether you want to build it yourself, learn with guidance, or have us build it for you — there's a path that fits where you are right now.

Free
DIY
$0
Free tutorial + open-source stack
Everything you need to build your Human 3.0 stack yourself. Step-by-step tutorial, template TELOS file, and all the setup instructions. We teach this at local libraries — it's designed to be accessible.
  • Full written tutorial (this site)
  • TELOS starter template
  • FRAME YAML templates
  • PAI setup instructions
  • Memory Palace Docker config
  • Community Q&A
Start the Tutorial →
Premium
Done-For-You
$1,997
Full build + 90-day support
We build your complete Human 3.0 stack from scratch — including dedicated hardware setup, custom AI persona, full TELOS development, and ongoing support for 90 days as you grow into it.
  • Everything in Done-With-You
  • Dedicated hardware setup (Mini PC or VPS)
  • Custom AI persona with your voice
  • Full TELOS written with you over 2 sessions
  • OpenClaw content engine configured
  • Custom skills built for your workflow
  • 90-day support + monthly check-ins
  • First right of access to team PAI Hub (when available)
Get Started →

🏛️ We Teach This at the Library — For Free

Every month, Gray Hodge teaches free hands-on workshops at local public libraries showing anyone how to build their own Personal AI using entirely free tools. No subscription required. No technical background needed.

Read the Free Tutorial → Bring a Workshop to Your Library