Forge Foundations v0.2

Forge Foundations

Primitive 001 - Ontology-Trained Forward Pass.

Forge Foundations is the primitive layer of the Forge Intelligence Engine. Primitive 001 now generates its classifier from the public ontology itself: nodes, titles, summaries, tags, routes, and relationships become a deterministic feature matrix before any larger AI model is introduced. The prediction is not magic. It is the result of visible weighted relationships.

Run Primitive 001 View Primitive Pipeline

Primitive

001 - Ontology-Trained Forward Pass

No AI API

No external model call

Weight Matrix

Generated from ontology records

Truth artifact

Evidence before reasoning

Primitive Construction Cycle

Intelligence is assembled step by step.

Primitive 001 turns text into a knowledge object: metadata, ontology signals, weighted scores, prediction, artifact, and relationship path. Each correction becomes lineage for the next run.

Input Content

Extract Metadata

Score Ontology Nodes

Select Prediction

Create Truth Artifact

Discover Relationships

Enter Intelligence Network

Live Primitive Execution

Watch a node enter the intelligence network.

This version generates its local classifier from the current public ontology. The important part is not a generic label; it is the forward pass from input tokens to weighted ontology fields, candidate node scores, final prediction, and inspectable truth artifact. We do not hide the mechanism. Forge shows the path from signal to evidence to relationship.

Forge Foundations v0.2

Primitive 001 - Ontology-Trained Forward Pass

We do not hide the mechanism. Forge shows the path from signal to evidence to relationship.

Paste Content for Primitive

Training Loop

01Capture Input

02Score Evidence

03Review Prediction

04Review / Train

05Training Trace

Ontology Forward Pass

A deterministic scoring pass from input tokens to ontology node prediction.

Tokenizing Input

Loading Ontology Profiles

Applying the Weight Matrix

Ranking Candidate Nodes

Selecting Prediction

Creating Truth Artifact

Entering the Intelligence Network

Current Step Output

Click Start, then advance one step at a time.

Ready

Start the primitive when you want to inspect the mechanism.

Nothing advances automatically. Click Run Primitive 001 or Start, then use Next Step to move through the classifier deliberately.

This panel shows only the output for the current step. Use the main Ontology Forward Pass controls above to advance the primitive.

Current Step Output

Click Start, then advance one step at a time.

Inspection Locked

The inspection surfaces unlock as the primitive creates evidence. This keeps the training loop from showing graph, artifact, and correction panels before they exist.

GraphStep 05 unlocks after prediction

ArtifactStep 06 unlocks after artifact creation

TrainingStep 05 unlocks when a prediction can be reviewed

The Double Helix

Fast intelligence needs governed evidence.

Fast intelligence is useful only when it is governed by traceable evidence.

Helix 1: Fast Classification

Ontology signals, weighted scoring, and small primitives route content quickly.

Helix 2: Governed Verification

Ontology rules, confidence thresholds, corrections, provenance, and evidence chains verify the result.

Truth Artifact

Classifications become inspectable records.

Every classification becomes a truth artifact: input, prediction, score, confidence, method, evidence, graph path, correction history, and lineage.

{
  "nodeId": "content_001",
  "input": "The ontology page should connect graph relationships, public context, memory, and the Forge Intelligence Network.",
  "predictedOntologyNode": {
    "id": "CTN-WEB-ONTOLOGY-INDEX-001",
    "title": "Public Ontology Index",
    "type": "web"
  },
  "score": 40.96,
  "confidence": 0.86,
  "method": "deterministic_ontology_forward_pass",
  "matchedTokens": [
    "Ontology",
    "Graph",
    "Relationship",
    "Public",
    "Memory"
  ],
  "evidence": [
    {
      "signal": "Ontology",
      "field": "Page Route",
      "weight": 4
    },
    {
      "signal": "Graph",
      "field": "Relationship",
      "weight": 2
    },
    {
      "signal": "Memory",
      "field": "Summary",
      "weight": 2
    }
  ],
  "graphPath": [
    {
      "source": "content_001",
      "relationship": "classified_as",
      "target": "Public Ontology Index"
    },
    {
      "source": "Public Ontology Index",
      "relationship": "projects_as",
      "target": "Website Ontology"
    }
  ],
  "correctionHistory": [],
  "createdAt": "timestamp"
}

Forge Foundations Catalog

Primitive 001 is the first of many.

The ontology-trained classifier is not the goal. It is the first reusable building block in a larger catalog of Forge Intelligence Engine primitives.

v0.1

Static Prototype Surface

Client-side deterministic prototype, visible evidence display, truth artifact preview, and local correction state.

v0.2

Ontology-Trained Forward Pass

Generate the classifier from public ontology records: nodes, titles, summaries, tags, routes, relationship labels, and related nodes become a weighted feature matrix.

v0.3

Human Correction Loop

Add correction modal, review notes, accuracy metrics, confidence distribution, and training lineage for every correction.

v0.4

Local Persistence

Persist truth artifacts, correction history, scoring traces, and graph-path records locally before promoting the primitive into a service.

v0.5

Internal Classifier Service

Expose Primitive 001 as an internal service consumed by the website ontology, public ontology, graph traversal tools, request intake, and future Forge systems.

v1.0

Forge Primitive 001

Graduate the ontology-trained classifier into a reusable Forge Intelligence Engine primitive with inspection, correction, lineage, and service boundaries.

Public Interface

Classify Content

Predict Ontology Node

Rank Candidate Nodes

Return Evidence

Return Confidence

Generate Truth Artifact

Local Infrastructure Path

01Browser

02Next.js

03Public Ontology

04Ontology Feature Matrix

05Weighted Forward Pass

06Candidate Node Ranking

07Ontology

08Truth Artifacts

09Relationship Engine

Forge Intelligence Engine

Primitive 001

Ontology-Trained Forward Pass

Active

Primitive 002

Text + Metadata Extraction

Planned

Primitive 003

Truth Artifact Generation

Planned

Primitive 004

Relationship Discovery

Planned

Primitive 005

Contradiction Detection

Planned

Primitive 006

Gap Detection

Planned

Primitive 007

Reasoning

Planned

Primitive 008

Workflow Recommendation

Planned

Primitive 009

Agent Orchestration

Planned

Primitive 100+

Planned

Graph Traversal Result

The smallest useful unit of Forge intelligence.

Before Forge routes work through large models, agents, or hosted GPU infrastructure, it must prove that local intelligence can understand the user's world.

Role

Forge Foundations proves the first Forge primitive: classify content, preserve evidence, accept correction, and improve.

Inherited From

Linus Torvalds for modular engineering, W. Edwards Deming for feedback, and Douglas Engelbart for human augmentation.

Implemented As

Local classifier prototype, truth artifact pattern, correction loop, and ontology routing primitive.

Visible Throughout

Forge Intelligence Engine, Public Ontology, Intelligence Stack, Website Graph, and future training lineage.

How This Fits in CTN

Graph traversal result.

Forge Foundations is a web node in the ConservaTech public ontology. The primitive layer of the Forge Intelligence Engine, beginning with local-first domain classification, truth artifacts, relationship discovery, and memory preservation.

Public Surface/forge-foundations

Forge Foundations is a web node in the ConservaTech public ontology with 12 explicit relationships. Forge Foundations does not yet have an explicit lineage source. Direct implementation links have not yet been modeled. Visitors can inspect it through /forge-foundations.

View Full Intelligence Graph