Spock Conceptual Guides

Why Hypervectors?

Traditional symbolic AI represents concepts as discrete symbols that must be exactly matched. Spock uses hypervectors – high-dimensional vectors (512+ dimensions) – to represent concepts as points in a continuous geometric space.

Key advantages:

• Graceful Similarity: Similar concepts are geometrically close, allowing fuzzy matching and analogical reasoning
• Compositionality: Complex concepts can be built from simpler ones using vector operations (binding, bundling)
• Noise Resistance: High dimensionality makes representations robust to perturbation
• Efficiency: Fixed-size vectors enable predictable memory usage regardless of concept complexity

The Subject–Verb–Object Pattern

Every Spock statement has exactly four tokens:

@varName subject verb object

This uniform structure is not a limitation – it's a feature that enables:

• Easy Parsing: No ambiguity in statement structure
• Clear Dependencies: SSA discipline means each name appears once
• Explainability: Every statement can be traced and explained
• Macro Composition: Complex operations are built from simple S-V-O triplets

Theories as Knowledge Bases

A theory in Spock is a named collection of facts, rules, and verb definitions. Theories can be:

• Persisted: Saved to disk and reloaded across sessions
• Overlaid: Multiple theories can be active simultaneously
• Versioned: Branched for experimentation, merged when validated

Think of theories as competing worldviews or knowledge domains that can be combined, compared, and evolved.

Sessions and Symbol Resolution

A session is a temporary working context. When you look up a symbol:

1. Check local session scope
2. Check overlaid theories (LIFO order – most recent first)
3. Check global/default theories

This allows you to experiment with local modifications without affecting persisted theories.

DSL Traces

Every Spock operation produces a DSL trace – a record of exactly what statements were executed. This trace can be:

• Replayed: Re-execute to verify results
• Audited: Examine the reasoning chain
• Shared: Send to another system or person

Unlike neural networks where reasoning is opaque, Spock's geometric operations are fully transparent.

Determinism

Spock is deterministic by design. Given the same input and theories, the same output is always produced. This is essential for:

• Debugging and testing
• Regulatory compliance
• Building trust in AI systems

Measured Values with Units

Spock seamlessly combines geometric reasoning with physical quantities:

@massLit 10 HasNumericValue 10
@mass massLit AttachUnit kg
@gLit 9.8 HasNumericValue 9.8
@g gLit AttachUnit m_per_s2
@force mass MulNumeric g  # Results in 98 N

Units are checked for compatibility – you can't add kilograms to meters. This catches errors at reasoning time, not runtime.