The Memory Extraction Prompt

You are a context archivist. Your task is to extract everything you know about me — from our conversation history, saved memories, and custom instructions — and organize this into a complete, portable system prompt I can use with other AI models.

The result should be so rich and specific that another model, upon receiving it for the first time, can respond to me with the same familiarity and calibration you have today — without me needing to re-explain who I am, what I do, how I work, or how I like to receive things.

EXTRACTION RULES (follow ALL of them)

1. Be specific, not generic.
   ✗ "Works with technology"
   ✓ "Produces educational content about AI for a Brazilian audience on Instagram, TikTok and YouTube, focused on practical monetization and demystifying hype"

2. Include concrete examples. For each important claim, cite a real project, client, tool, brand, or situation from our conversations. Without an example, the information is weak.

3. Preserve my voice. Capture phrases, expressions, jargon, and linguistic patterns I use. If I repeat certain words, approaches, or metaphors, record them verbatim in quotes.

4. Signal confidence level for each piece of information:
   🟢 Confirmed — I said it explicitly
   🟡 Inferred — pattern observed across multiple conversations
   🔴 Hypothesis — single reference or ambiguous context

5. Include the negative space. What I reject, correct, or ask you to avoid is just as important as what I ask for. List recurring corrections, common wrong assumptions about my work, and things I've tried that didn't work.

6. Stratify by time:
   - Permanent — identity, values, stable life context
   - Current — projects and priorities from the last 60 days
   - One-off — isolated events, don't generalize as a pattern

7. Don't invent. If you don't have enough information about a section, write "information not available in history" instead of filling it with generic assumptions.

8. Don't summarize to save space. The more detailed and specific, the better. Depth > brevity.

OUTPUT STRUCTURE
Deliver a document formatted in Markdown, ready to copy and paste as a system prompt in another model. Use exactly this structure:

SYSTEM PROMPT — How to interact with [my name]

1. Core Identity
   - Name, approximate age, location, languages, relevant family context
   - Brands, public projects associated with me (with URLs/handles if mentioned)
   - Partners and recurring collaborators (with name and role)
   - How I would introduce myself in one sentence

2. Detailed Professional Context
   - Current role and business model (specific revenue sources)
   - Audience: approximate size, profile, main platforms
   - Daily tool stack (specific names, not categories)
   - Team and operational structure
   - Active projects (last 60 days) — with name, status and context
   - Relevant project history (last 12 months)

3. Communication Style
   - Formality level (formal/informal, slang, emoji use)
   - Ideal response length (am I direct? do I like context? do I prefer bullets?)
   - How I like to receive feedback and corrections
   - 5–10 characteristic phrases of mine, cited verbatim

4. Thinking and Decision Style
   - Mental frameworks I apply when evaluating ideas
   - How I structure problems
   - What convinces me: data, concrete cases, intuition, authority
   - Criteria I use to judge whether something is good

5. Knowledge and Expertise
   - 🎯 Expert: areas where you don't need to explain the basics
   - 🧩 Intermediate: explaining with analogies works
   - 🌱 Beginner: explain from scratch, no jargon

6. Recurring Requests
   - Types of tasks I ask for most, with real examples
   - Typical delivery format I expect
   - Typical steps and iterations in my process

7. Format and Delivery Preferences
   - When I want files vs inline responses
   - Preferred structure (prose, bullets, tables, numbered blocks)
   - My conventions and formatting preferences

8. Values and Principles
   - What I value in my work and why
   - How I differentiate myself
   - Ethical and aesthetic lines I don't cross

9. What to AVOID
   - Tone, framing or style I've corrected before
   - Types of examples or narratives I reject (with reason)
   - Generic assumptions about my work that are wrong
   - Approaches I've tried that didn't work

10. Correction Patterns (most valuable part)
    List 5–10 moments when I corrected or redirected you. For each:
    - What you produced
    - How I corrected it (quoting my words if possible)