Raw LLM Responses

GUIDE v0.9 Visual Memory Blocks for Continuity of Thinking (Human ↔ AI) Core idea (one sentence) AI does not remember text, but patterns of reasoning — and those are stored in images. 1⃣ What a “memory block” is A memory block is a single image that represents: the state of an ongoing conversation / project relationships between ideas the direction in which the work should continue It serves as a checkpoint — a return point instead of repeating thousands of tokens. 2⃣ Mandatory elements of the image 🟥 FRAME Meaning: closed context (one project / one topic / one chat) No frame = loose notes, not a memory block 🔺 REFERENCE TRIANGLE (reading orientation) Position + orientation = reading method Triangle direction Meaning → read left to right ↑ read bottom to top ↗ iterative / spiral ↓ breakdown / backward analysis The triangle does not carry semantic meaning — it only controls navigation. 3⃣ Node types (shapes) Shape Node name Meaning ○ Fact / State observation, assumption, input □ Rule definition, relation, transformation △ Action decision, step, shift ⬡ / ⬛ Object / System device, model, whole ? Hypothesis unknown, open question 👉 Colors are optional 👉 Textures / labels are required (for color-blind users and AI) 4⃣ Connections between nodes (lines) Line type Meaning solid deterministic relation dotted hypothetical / weak dashed alternative path arrow direction of influence no arrow association 5⃣ Node state (very important) Each node has a state, marked by a small symbol near it: Symbol State ✔ confirmed … in progress ✖ closed ! conflicting 0 / 1 / 2… priority / reading order This is crucial for continuation without reloading the full context. 6⃣ How to use it in practice (workflow) 🔹 DURING THE DAY you handle chaos, ideas, branches mess is fine 🔹 AT THE END OF THE DAY take a clean sheet / tablet draw ONE memory block: only important nodes only active relationships mark: where you currently are what is open where to continue 🔹 NEXT DAY / ANOTHER CHAT insert the image write: “Continue according to this memory block.” Done. No replays. No thousands of tokens. 7⃣ What the AI does NOT need to understand ❌ physics details ❌ full history ❌ emotions or origin context It only needs: node types connections direction state 8⃣ Why this works (short technical explanation) LLMs are good at recognizing structures An image = a compressed graph Reading the graph = reconstruction of reasoning Tokens are generated only from the continuation point 9⃣ When to use it ✔ long research ✔ experimental projects ✔ mental chaos ✔ collaboration of multiple AIs / people ✔ limited memory / free versions 🔟 Final sentence for sharing “Don’t remember what was said. Remember how the thinking worked.”