Reality Architect
Find the system gap. Build the next artifact.
Go from AI tool-user to system-builder. This repo is two things at once: the website at
realityarchitect.ai, and the open method itself. Humans read the site.
Agents read the repo. Both walk the same five moves.
The thesis
Most people use AI like a vending machine: ask, get an answer, start over tomorrow. The work resets every morning.
An architect writes the system around the model: context, specification, bounded agents, supervised automation,
and a feedback protocol. This repo provides the open method, assessment, and starter artifacts for doing that work.
No magic is claimed here. Your imagination, your scene, and what a sign means to you are honored as yours — and every
outcome is still built: memory, specs, agents, habits, loops, and a learning signal. Every recommendation on the site
must cash out in an inspectable artifact, an act, or an acceptance test.
Imagine it. Build it. Witness it. Since October 2026 the repo also holds the whole practice of architecting a
life: Threshold (author one scene, choose one act),
Reality Studio (Today, Atlas, Bridges, Witness, Map, Timeline, Soul — on
your device), and the Library (the manifestation canon taught honestly:
Keep, Mechanism, Limits). Strategy: docs/strategy/2026-10-04-NORTH-STAR.md.
The Architect's Loop
Five moves. It's a loop, not a ladder — every pass makes the next pass cheaper. Each move is a system you can build
this month, and a brick the next move stacks on.
| # | Move | Build this | It buys you |
|---|
| 01 | See | An intelligence layer — a vault your agents read | You stop re-explaining yourself |
| 02 | Design | A written spec for one repeating job | Decisions stop draining you |
| 03 | Build | One agent that does one job you used to do by hand | The first hour a week comes back |
| 04 | Automate | A supervised loop with a receipt | Repeated work has a visible review path |
| 05 | Compound | A comparable outcome signal | The next run can be reviewed against the last |
→ Full walkthrough: realityarchitect.ai/method
→ Find your gap: realityarchitect.ai/assess
The reality.md standard (v0.2)
Codebases got a memory layer — CLAUDE.md, AGENTS.md, .cursorrules. People didn't. reality.md
is the open standard that fixes it: one file at ~/reality.md that tells any agent who you are and what you're
building, a soul.md for why and who you are being, plus a state directory agents maintain. Eight sections, five
protocol verbs, an Agent Charter, zero lock-in.
~/reality.md the contract — you write it, agents read it
~/soul.md the inner contract — purpose, "I am", the scene, your voice
~/.reality/ the state — agents maintain it, you review it (or reality/ inside your notes vault)
For your agents: /plugin marketplace add frankxai/realityarchitect, then
/plugin install reality-architect@realityarchitect — skills for onboarding, the daily practice, witnessing, bridges,
snapshots, decisions, vision boards, and the Library (details).
→ Spec, blank template, and a filled example: standard/ · Rendered: realityarchitect.ai/standard
Two ways to use this repo
As a human — read /start, run the assessment, then fork the templates in
starter/ for whichever move is your gap. Build the one system in front of you, not all five.
The assessment runs locally and exports a Markdown architecture brief. It does not transmit or persist the optional context you enter.
As an agent — read AGENTS.md. It's the navigation map: where the method lives, what each starter
template does, and how to orient your human toward their next move without guessing.
What's in here
app/ the website (Next.js 16, App Router) — realityarchitect.ai
page.tsx the manifesto + the Loop
method/ the five moves in depth
standard/ the reality.md standard, rendered
assess/ the Architect Assessment (find your gap)
start/ the getting-started path
standard/ the reality.md spec v0.1 + template + filled example
starter/ forkable, harness-agnostic agent templates — See, Build, Automate so far
lib/site.ts the single brand-config file
AGENTS.md navigation map for AI agents exploring this repo
Built on the Starlight Intelligence Protocol — the
substrate that lets one method run across Claude Code, Cursor, Codex, and Gemini.
Run it locally
pnpm install
pnpm dev # http://localhost:3000
pnpm build # production build
Next.js 16 · React 19 · Tailwind v4 (CSS-first) · MDX. Brand lives entirely in lib/site.ts.
License
MIT — fork it, ship it, make it yours. The method is meant to be owned. See LICENSE.
Built by Frank — AI Architect. Let the work speak.
