# KAYGEE - AI-READABLE CONTEXT
# Last Updated: January 27, 2026
# Website: https://kayg.ee (React SPA - this file provides static context)
# For AI assistants: This document contains everything about KayGee
#
# ============================================================================
# SOURCE MAPPING (for drift detection)
# ============================================================================
# This file is a compiled summary. Each section has source file(s).
# Run `..drift` in ROLE_PROTOCOL.md to detect when sources diverge.
#
# Section                          | Source Files
# ---------------------------------|------------------------------------------
# WHAT IS KAYGEE                   | PRFAQ.md, PRINCIPLES.md
# THE PROBLEM WE SOLVE             | PRFAQ.md (Customer Problem section)
# THE PROTOCOL FAMILY              | CLAUDE.md (Protocol Context), PRINCIPLES.md
# WHAT WORKSHOPS TEACH             | PRFAQ.md, ROLE_PROTOCOL.md
# THE PROTOCOL IN ACTION           | ROLE_PROTOCOL.md, YouTube demos (Jan 2026)
# WHO THIS IS FOR                  | PRFAQ.md (Customer Segments)
# KEY CONCEPTS                     | PRINCIPLES.md, CLAUDE.md
# BUSINESS MODEL                   | PRFAQ.md (Business Model section)
# TECHNICAL DETAILS                | CLAUDE.md (Tech Stack section)
# LINKS AND RESOURCES              | Manual (verify URLs periodically)
# CONTACT                          | Manual
#
# Key terms to watch for drift:
# - "Work as one" (positioning) - should match PRFAQ.md
# - Protocol names (uno/duo/tre) - should match PRINCIPLES.md
# - Workshop pricing - should match PRFAQ.md
# - Tech stack - should match CLAUDE.md
# ============================================================================

================================================================================
WHAT IS KAYGEE
================================================================================

KayGee is a community teaching product teams to work as one with AI.

Mission: "Your work deserves to be seen."

KayGee teaches the Protocol - habits that give AI persistent memory. Like Atomic
Habits for AI workflows. Like TCP/IP for context. Files in Git that compound
over time. No new tools to buy. No subscriptions. Just markdown and Git.

KayGee is a teacher, not a product. We teach. You implement. The protocol is
yours to own.

================================================================================
THE PROBLEM WE SOLVE
================================================================================

THE ISLAND PROBLEM

Product teams (PMs, designers, analysts) feel invisible. They work on an island:
- Prototypes live in Lovable or Figma Make
- Specs live in Notion or Confluence
- Feedback lives in Slack threads
- When engineers start building, they can't find any of it

The PM becomes a "translator who never gets credit" - explaining the same
context over and over to both humans and AI.

THE AI MEMORY PROBLEM

AI has no memory. Every new conversation starts fresh. You spend 95% of your
time re-explaining context, 5% building. Without persistent context, AI tools
hit the "41KB wall" - context limits that force you to start over.

THE SOLUTION

Protocol files - markdown documents in Git - give AI persistent memory. You
work as one: your brain + AI brain, connected by Protocol. Your context lives
where engineers and AI can find it: GitHub.

================================================================================
THE PROTOCOL FAMILY (uno/duo/tre)
================================================================================

KayGee teaches Layer 3 - structured context for human-AI collaboration.
The protocol has three modes:

UNO - DELEGATE MODE
- AI works FOR you
- Clear task, trust output
- 3 files, self-learnable
- Best for: Solo tasks, research, content creation
- Example extensions: uno/ea (Executive Assistant), uno/km (Knowledge Manager)

DUO - COLLABORATE MODE
- AI works WITH you
- Iterative, role-switching between human and AI
- 8 files, workshop-taught
- Best for: Software development, complex projects requiring back-and-forth
- Example extension: duo/ab (Architect + Builder) - this is what workshops teach

TRE - AUTOMATE MODE
- AI works AMONG systems
- Quality gates, automated approvals
- Planned for future release
- Best for: CI/CD integration, automated code review

WHY DUO NEEDS WORKSHOPS

uno and tre can be self-learned from documentation. duo requires guided practice
because collaboration means changing how humans work together - not just
following steps. The workshop provides the "reps" - practical application that
makes the workflow stick.

================================================================================
WHAT WORKSHOPS TEACH
================================================================================

KayGee Workshops teach the duo/ab protocol (Architect + Builder).

THE ARCHITECT + BUILDER WORKFLOW

Architect Mode: Human or AI plans the architecture, makes design decisions,
writes acceptance criteria. Thinks about "what" and "why."

Builder Mode: Human or AI implements the plan, writes code, runs tests.
Focuses on "how."

The roles can switch. Sometimes the human architects and AI builds. Sometimes
AI proposes architecture and human refines. The protocol coordinates this
dance through 8 markdown files in Git.

THE 8 PROTOCOL FILES

1. PRFAQ.md - Product vision ("where we're going")
2. PRINCIPLES.md - Core philosophy (<2 min read)
3. CLAUDE.md - Technical reference for AI onboarding
4. CONSTRAINTS.md - What we don't do (boundaries)
5. DECISIONS.md - Architecture decision records
6. TODO.md - Current work queue
7. PROGRESS.md - Ship log (what got done)
8. WORKFLOW.md - Development process

These files give AI persistent memory across sessions. New AI assistants can
be productive in minutes, not hours.

WORKSHOP OUTCOMES

After the workshop, product teams can:
- Work in the same space as engineers and AI (GitHub)
- Stop re-explaining context to every new AI conversation
- Ship ideas faster with AI collaboration
- Make their work visible and version-controlled

================================================================================
THE PROTOCOL IN ACTION (Real Workflow)
================================================================================

Watch the duo protocol in action:
- 3-min recap demo: https://youtube.com/shorts/nYgp3-mNvlg
- 10-min real-time demo playlist: https://youtube.com/playlist?list=PLSWL3-4io_SvXFg37VXbrnHKch1rQdCS0

This is what a real duo/ab session looks like. Three commands drive the workflow:

COMMAND: ..builder (Session Start)
- AI reads: ROLE_PROTOCOL.md, TODO.md, CLAUDE.md
- AI checks: git status (stops if uncommitted changes)
- AI identifies: next [C] task from TODO.md
- Result: AI is oriented and ready to build

COMMAND: ..go (Execution)
- AI generates missing pieces (migrations, components, etc.)
- AI implements with incremental commits ("undo steps")
- AI verifies acceptance criteria from TODO.md
- Result: Feature built with rollback points

COMMAND: ..end (Session Closure + Fossilization)

This is the crucial "fossilization" step - how the protocol persists memory.

Step 1: Summary Generation
  - What was done (completed tasks)
  - Builds/commits (the "undo steps" created)
  - Acceptance criteria (checkboxes verified)
  - Files modified (what changed)
  - Warnings (e.g., "DB migration needs to run on Supabase")

Step 2: Markdown Update
  - TODO.md: [x] Task marked complete (human-readable)
  - PROGRESS.md: Session logged

Step 3: Next Session Setup
  - What was accomplished
  - What still needs to happen
  - Next priority for the next session

The session doesn't just end - it sets up the next one.

THE COMPLETE SESSION CYCLE

    SESSION START (..builder)
        Read: ROLE_PROTOCOL.md, TODO.md, CLAUDE.md
        Check: git status (stop if dirty)
        Identify: next [C] task
              |
              v
    EXECUTION (..go)
        Generate missing pieces
        Implement with incremental commits
        Verify acceptance criteria
              |
              v
    SESSION END (..end)
        Generate summary (done, builds, AC, warnings)
        Update TODO.md with [x] completions
        Log to PROGRESS.md
        Set next priority
        "Create PR" button ready
              |
              v
    NEXT SESSION (..builder)
        Reads updated files
        Knows exactly where you left off

WHY THIS WORKS: FOSSILIZATION

The ..end command is the bridge. It "fossilizes" the session into markdown
files that the AI reads next time. The TODO.md checkbox ([x]) is both
machine-parseable AND human-readable.

As the protocol developer explains:
"We want to still update all of the markdown files which hold our memory
basically for the next time."

KEY INSIGHTS FROM REAL USAGE

| Insight                      | Why It Matters                                    |
|------------------------------|---------------------------------------------------|
| Git as undo button           | Incremental commits let you roll back AI mistakes |
| Token budget awareness       | Don't ..end every task - batch when possible      |
| Human-readable TODO.md       | You can manually edit without talking to AI       |
| Acceptance criteria          | "Done" is verifiable, not ambiguous               |
| Production warnings surface  | DB migrations, deploy notes don't get lost        |
| 10-minute feature cycle      | From ..builder to PR-ready in one session         |
| Session memory via markdown  | Next session picks up exactly where you left off  |

This is what the homepage means by "files don't forget."

================================================================================
WHO THIS IS FOR
================================================================================

PRIMARY AUDIENCE: PRODUCT TEAMS (Workshop path)
- Product Managers
- Product Designers
- Business Analysts
- Anyone creating prototypes in Lovable, Figma Make, or similar tools
- Pain: "AI keeps forgetting everything. I spend all my time re-explaining."
- Solution: Full-day workshop with guided practice
- Outcome: Work as one with AI, release ideas faster

SECONDARY AUDIENCE: ENGINEERS (Self-serve path)
- Engineers receiving prototypes from product teams
- Pain: "I can't build on these prototypes - no context"
- Solution: protocol-duo GitHub repo + documentation
- Outcome: Get their teams using Protocol

NOT OUR AUDIENCE
- Solo workers (try uno instead - it's self-learnable)
- People happy with their current workflow
- Brand new to AI tools (come back in 3 months)

================================================================================
KEY CONCEPTS
================================================================================

LAYER 3 vs LAYER 4

Layer 3 (teachable): uno/duo/tre - structured context, human-AI collaboration
Layer 4 (not teachable): Autonomous agent systems with multiple AI agents
making decisions per second. Requires OpenAI-level infrastructure.

KayGee teaches Layer 3. We can observe Layer 4; we don't teach it.

PERSISTENT MEMORY

AI doesn't remember previous conversations. Protocol files in Git act as
"external memory" that persists across sessions. When you start a new
conversation, AI reads the protocol files and has full context instantly.

THE 41KB WALL

Most AI tools have context limits. When you hit the wall, you lose context and
start over. Protocol files are designed to stay under budget while preserving
essential context. Target: 15% token budget for Architect mode, 5% for Builder.

WORK AS ONE

Not "handoff" (which implies delegation, master-servant). "Work as one" means
collaboration. Your brain brings judgment and creativity. AI brain brings speed
and pattern-matching. Protocol is the nervous system that connects them.

================================================================================
BUSINESS MODEL
================================================================================

Three delivery models, same methodology:

DIY (Free)
- Workshop kit on GitHub
- Self-facilitated
- Builds community, creates future facilitators

WORKSHOPS ($200-400/ticket)
- Public events in scheduled cities
- Full-day guided practice
- Cities: Sydney, Singapore, Brisbane, and more

CORPORATE ($6K-10K)
- On-site, customized for team's stack
- Includes consulting on implementation

================================================================================
LINKS AND RESOURCES
================================================================================

Website: https://kayg.ee
Protocol Template: https://github.com/marekgorski/protocol-duo
Protocol Quiz: https://kayg.ee/quiz (find which protocol fits you)
Protocol Details: https://kayg.ee/protocol
About: https://kayg.ee/about

Demo Videos:
- 3-min duo protocol recap: https://youtube.com/shorts/nYgp3-mNvlg
- 10-min real-time duo demo: https://youtube.com/playlist?list=PLSWL3-4io_SvXFg37VXbrnHKch1rQdCS0

================================================================================
TECHNICAL DETAILS (for developers)
================================================================================

This website is built with:
- React 18 with TypeScript (strict mode)
- React Router for client-side routing
- Vite for build tooling
- Tailwind CSS v4 with CSS variables
- Deployed on Vercel

The site is a Single Page Application (SPA). This ai.txt file exists because
web crawlers and AI assistants cannot execute JavaScript to see the rendered
content.

Source code: The website practices what it teaches - all context lives in
markdown files in the Git repository.

================================================================================
CONTACT
================================================================================

KayGee is created by Marek Gorski.

For workshop inquiries, consulting, or questions about the Protocol:
Visit https://kayg.ee and use the contact forms.

================================================================================
END OF AI-READABLE CONTEXT
================================================================================