Session 3 — Knowledge Landscape & .nci Format

What We Built

The top-level brain object that owns everything — KnowledgeLandscape. Added consolidation (nodes mature over time), inter-cluster connections (pathways across the brain), and the .nci binary file format for saving and loading a complete brain.

---

Key Concepts

Knowledge Landscape

The skull that holds the brain. Owns all nodes, clusters, inter-cluster connections, and metadata. Built in one call via landscape.build(). Handles querying with auto-reinforcement, consolidation cycles, and save/load.

Analogy

The skull that holds the brain. Neurons (nodes) and brain regions (clusters) live inside it. It manages the whole system as one unit.

Consolidation

A score from 0.0 (fresh) to 1.0 (crystallized) on every node. Reinforced nodes consolidate — their threshold drops and they resist decay. Unused nodes decay back toward fresh.

Analogy

Learning to drive. Day 1: every action is conscious (fresh, high threshold). Year 10: it's automatic (crystallized, low threshold). Stop driving for a year and you're rusty — but it comes back fast.

.nci File Format

Custom binary format for saving and loading a complete Knowledge Landscape. Header with magic bytes "NCI!" and version, then packed numpy arrays for all node data, JSON for cluster metadata, and the LSH projection matrix.

Analogy

Freezing a brain and thawing it later with all memories intact. The .nci file IS the brain on disk.

Inter-Cluster Connections

Clusters connect to their most similar neighbors automatically. Creates pathways across the landscape — networking → security → linux. Built by comparing cluster centroids.

Analogy

Hallways between sections of a library. The networking section has a door to the security section, which has a door to the Linux section.

---

Design Principle — Nodes Never Disappear

Core Design Decision

Nodes are never removed through decay. Decay is priority sorting only — unused knowledge becomes harder to activate but never disappears. Nodes are only removed by explicit deprecation with a logged reason.

---

Benchmarks Achieved

Metric	Target	Result
.nci save (100k nodes)	< 5 sec	0.4s ✅
.nci load (100k nodes)	< 5 sec	1.6s ✅
File size (100k nodes)	—	~34 MB
Consolidation	reinforced > fresh	0.50→0.70 ✅
Inter-cluster connections	> 0 per cluster	3.0 avg ✅

---

Key Files

• index.py — KnowledgeLandscape, consolidation, save/load

Key Methods

landscape.build(num_nodes, num_clusters)  # build complete brain
landscape.query(signature, top_k)         # query with reinforcement
landscape.consolidation_cycle()           # decay unused nodes
landscape.save("brain.nci")               # save to binary file
KnowledgeLandscape.load("brain.nci")      # load from binary file