Video summary

Ce dossier rend ton IA 10x plus pertinente (mon setup complet 2026)

Main summary

Key takeaways

Technology

Core idea (problem → solution)

  • Problem with most AI chat setups (Claude/ChatGPT/Cloud Code, etc.): each new conversation/session starts from scratch, so you repeatedly re-enter the same brand/background/preferences, formulas, workflows, and “credits.” The assistant also forgets what it learned previously—unless that info is manually included in that specific project.
  • Impact: you end up switching conversations often and wasting time (he estimates ~30 minutes/day, compounding over a year).
  • Solution: create a persistent “brain”—a structured folder of knowledge and instructions—so the AI can re-use your context across tools and over time, and essentially self-improve by learning what to store.

The “Brain” concept / product feature

  • The “brain” is a single structured folder on your computer (optionally on USB, readable and portable).
  • The folder contains Markdown (MD) files, because the claim is that AIs read MD well and it’s efficient for context usage.
  • Designed to be future-proof:
    • AI tools/models will change, but your data stays the same.
    • You can plug the folder into whichever AI you use next without redoing everything.

Why it’s “non-negotiable” (analysis claims)

He argues this is the one setup most people skip. Claimed main benefits include:

  • Zero repetition: the AI remembers your preferences, style, and past decisions/errors.
  • Cross-referencing: tasks like offer/campaign analysis, order comparisons, and strategy outputs that would normally take hours can be done faster (he suggests “in 2 minutes”).
  • Continuous improvement: the system logs mistakes and “lessons learned” so the AI “never makes the same mistake twice” (within the scope of what you record).

Architecture (folder structure)

He proposes a folder with subfolders such as:

  • Brain (root folder)
    • Body (personal identity/preferences/objectives)
    • Personality / Objective (who you are, your style, constraints)
    • Brand (product/people/prices/reviews/orders/creative assets)
    • Knowledge (training materials, exported tweets/YouTube/training notes, and conclusions/summaries)
    • Memory
      • Contains a cloud.md file used to manage memory.
  • Projects (active workflows)
  • Tools (scripts/skills, e.g., Python scripts)
  • Creative Lab (optional: creative production systems)
  • Resources (optional external stash like PDFs/transcripts, exported materials kept outside “Brain” so the AI doesn’t have to read everything every time; it pulls what’s relevant on demand)

He also stresses quality over quantity: export only information that’s genuinely high-quality to avoid mediocre output.

Workflow / tutorial (step-by-step)

  1. Create the folder structure (“Brain”, “Tools”, “Knowledge”, etc.) with MD files.
  2. Use a speech-to-text / assistant to interview/setup your preferences:
    • He recommends apps like Super Whisper (also mentions alternatives like WHIS Flow / “WHP Flow”).
    • You can speak, it transcribes into text, and the AI structures everything.
  3. Extract, summarize, organize:
    • Export relevant training, Discord server content, YouTube transcripts, Twitter, etc. into MD.
    • Configure “auto-learning” by setting up the root file (the cloud.md / memory system).
  4. Session context continuity:
    • At the end of each chat, the AI writes a brief for the next session, so continuity is preserved even if you switch conversations.
    • You can manually tell it “Remember this” to store key facts.
  5. Auto-management across AI platforms:
    • If you switch from one AI to another (e.g., Claude → GPT-style model), you just continue the task and the AI should find relevant context in your exported files.

Concrete examples of what the system enables (claims)

He describes having the AI able to:

  • Recall past work (e.g., “BTS we did a week ago”).
  • Re-export Shopify orders, then compare/assess performance (e.g., whether BTS was positive).
  • Pull relevant material from exported training sources (example: an Evolve training school, plus exported Discord server content).
  • Generate consistent outputs using reusable “skills” (briefs → structured skills).
  • Build/duplicate workflows:
    • Once he defines “creative briefs” matching his exact requirements, he can instruct the AI to create a skill, so future executions follow the same process reliably.

What to do with the video (suggested “how-to”)

  • Feed the entire transcript of the video to the AI.
  • Ask something like: “Set it up for me” so it initializes the brain/memory structure automatically.
  • He suggests he’ll provide setup details in comments if needed.

Main sources / speakers

  • Speaker: the video narrator/author (not otherwise named in the subtitles).
  • Referenced products/tools: Claude / ChatGPT (implied), Cloud Code, Super Whisper, WHP/WHIS Flow, plus general mention of GitHub tools for exporting Discord servers.

Original video