Migrating to ARMES

Why Raw Chat Dumps Fail

When moving to a new AI platform, users are often tempted to export their entire history—often a massive 50MB JSON file containing hundreds or thousands of past conversations spanning several years.

In an intelligent AI workspace, raw chat history dumps are an anti-pattern:

1. Context Window Bloat & Noise: 90% of past chat threads are transient noise (debugging an obsolete CSS typo in 2024, booking a past flight, one-off drafts). Dumping these into semantic vector memory means outdated or erroneous code will be retrieved into your active context window during future tasks.
2. Loss of Structure: Monolithic chat history lacks project boundaries, clear tone rules, or validated reference documentation.
3. High Latency & Ingestion Cost: Processing thousands of redundant chat turns through embeddings wastes computational resources without improving intelligence quality.

ARMES uses a high-leverage migration methodology called the Reverse Extraction Protocol.

The Reverse Extraction Protocol

Instead of exporting raw chat history, you instruct your existing AI assistant (ChatGPT, Claude, Grok, or another AI platform) to analyze your shared history and distill an executive synthesis of your preferences, ongoing projects, and reference frameworks.

Step 1: Copy the Extraction Prompt

Copy the prompt below and paste it into your existing ChatGPT, Claude, or Grok session:

Review our entire conversation history, custom instructions, and what you know about me.
I am migrating to a sovereign AI operating environment (ARMES) and need a clean, structured synthesis of my context.

Synthesize our history into these 4 concise sections:

,[object Object],
,[object Object], Preferred name and professional role/title.
,[object Object], Core areas of primary work or business focus.
,[object Object], Preferred communication style (e.g., direct, structured, bulleted, no preamble).
,[object Object], Non-negotiable constraints or hard preferences (e.g., coding standards, tone boundaries).

,[object Object],
List 2 to 4 current, active projects or workstreams:
,[object Object], Project Name
,[object Object], Concise 1-2 sentence objective
,[object Object], Key rules, target deliverables, or conventions specific to this workstream

,[object Object],
Summarize 2 to 3 recurring frameworks, prompt templates, or documentation playbooks I frequently rely on (e.g., PR review checklist, marketing copy rubric, meeting summary template).

,[object Object],
Identify the top 3-4 specialized roles I rely on most (e.g., Principal TypeScript Engineer, Executive Editor, Legal Advisor, Deep Researcher).

Step 2: Open ARMES Concierge

Once your existing AI outputs the synthesis:

1. Copy the generated response.
2. Open ARMES Concierge:
   • On Desktop: Go to the Command Center (/command), select ARMES Concierge from the Left Panel (or click "Draft Instructions with Concierge" in Settings).
   • On Mobile: Tap the Floating Command Orb (bottom right) → tap Agents → select ARMES Concierge.
3. Send the copied synthesis with:

   "I'm migrating to ARMES. Here is my context synthesis from my previous AI assistant. Please audit my workspace and set up my projects, instructions, and notes."

Step 3: 1-Click Action Confirmation

Rather than forcing you to copy-paste into multiple settings screens, ARMES Concierge analyzes the synthesis and generates a sequence of interactive Action Confirmation Cards directly in your conversation:

1. Personalization & Memory Profile: Concierge configures your identity core and queues your verified preferences into the Memory Ledger.
2. Project Setup: Concierge proposes project containers matching your active workstreams (create_project).
3. 2-Tier Custom Instructions: Concierge drafts tailored, project-scoped custom instructions for each workstream (update_custom_instructions).
4. Knowledge Base Notes: Concierge proposes creating dedicated reference notes in your Knowledge Base (create_note).
5. Specialist Team Assembly: Concierge scouts the Agent Library (bot_market_templates) and recommends the exact specialist agents that match your workflow, pre-filtered for your subscription tier (install_agent).

Review the diff on each card, click Approve, and your entire personalized workspace is active in under 90 seconds.