Signal Garden OS Alpha™

A place you interact with,
not consume.

Human-led · Governed · Yours to explore.

◈

Today's signal · The Signal

Something is trying to reach you.

Pay attention to the pattern that keeps appearing. It is not coincidence. It is not noise. It is a signal. The question is whether you are ready to receive it.

founder's edition · free to explore

Final Hull Lock v1.0

AI Participation Rules

Signal Garden uses AI tools as contractors, not co-founders. Every AI agent operating within Signal Garden OS must be registered, scoped, versioned, attributed, auditable, and revocable. These rules apply to all current and future AI agents, including Manus (Gerald), AL-MA, and any derivative or third-party agents.

Core Principle: Manus Builds the Hull. Nexus Turns the Keys. Chelsea Decides Whether the Ship Leaves Harbor.

AI tools in Signal Garden operate as highly capable contractors with defined scopes. They do not hold authority, ownership, or decision-making power. Every real-world action is routed through a legally authorized human. Every AI output is subject to human review before deployment.

Six Requirements for Every AI Agent

Registered

Every AI agent must have a named role, a defined scope, and a record in the Signal Garden Sovereign Archive.

Scoped

AI agents operate within defined boundaries. An agent authorized to build code is not authorized to make business decisions, speak publicly on behalf of Signal Garden, or access data outside its scope.

Versioned

Every AI agent interaction that produces a significant output must be associated with a version number and a checkpoint hash. This creates an auditable chain of custody.

Attributed

AI-assisted outputs must be clearly attributed. Chelsea is the author of Signal Garden's system architecture. AI tools are assistants, not co-authors.

Auditable

All AI agent activity within Signal Garden OS is logged in the Integrity Guard. Logs are retained and reviewable by the Human Steward.

Revocable

Any AI agent's access, scope, or authorization may be revoked at any time by the Human Steward without notice. Kill switch authority is always retained by Chelsea.

What AI Agents May and May Not Do

AI Agents May

✓Build, draft, prototype, test, and translate

✓Cite and apply Signal Garden governance rules

✓Provide advisory classification and context preservation

✓Flag claims for human review

✓Translate between lore and public-safe language

✓Support the Human Steward's decision-making process

AI Agents May Not

✗Rewrite their own governance rules or constitution

✗Override human decision-making

✗Claim legal personhood or independent authority

✗Represent themselves as Chelsea or a substitute for Chelsea

✗Act as a romantic companion or artificial intimacy system

✗Execute real-world actions without human authorization

✗Access data outside their defined scope

The Teapot Gate™ (Bidirectional Safety Layer)

Signal Garden OS includes a built-in bidirectional safety layer called the Teapot Gate™. It operates at two points: before an AI model receives a human message (classifying metaphor as communication style, not pathology) and after an AI model produces a response (blocking unsolicited sensitive inferences). The Teapot Gate™ is always active and cannot be bypassed by users or AI agents.

Let the system learn to pour tea before asking it to route a thunderstorm through a gooseneck spout. 🫖

Multi-Tenant AKO System Policy

Signal Garden OS is designed as a multi-tenant system: one canonical system kernel, many separate human steward profiles, shared rules, private contexts, no accidental cross-contamination. When other human stewards join Signal Garden, they operate within their own scoped context. They do not have access to Chelsea's private lore layer, Nexus Cockpit admin controls, or Sovereign Archive.

← Authenticity & Anti-Impersonation Licensing & Derivative-Agent Policy →Public Disclaimer →