Prompt Engineering Glossary
A living glossary of prompt engineering terms, updated periodically.
CORE CONCEPTS
Prompt Engineering – Crafting inputs to shape outputs predictably
Signal Density – Ratio of useful information to fluff; how much meaning per word
High-value Tokens – Words or phrases that strongly affect the model's interpretation and output
Semantic Compression – Expressing more meaning in fewer words without losing clarity
STRUCTURAL TECHNIQUES
Top-loading – Placing key information at the beginning where the model pays most attention
Weighting – Emphasizing certain elements more than others to guide priority
Order Bias – LLM tendency to prioritize earlier tokens in the input over later ones
Structured Output Specification – Defining the format or structure you want the output to take (e.g., JSON, markdown, React component)
CONTROL METHODS
Soft Control – Minimal specification that allows organic emergence while maintaining direction
Negative Prompting – Explicitly excluding or minimizing unwanted elements
Constraint Declaration – Stating limitations or boundaries upfront to focus the response
Tonal Anchoring – Using consistent voice or style markers to stabilize tone across outputs
Identity Anchors – Core personality traits or characteristics that define a character or voice
Context/Scene Grounding – Shaping behaviour and responses through environmental or situational framing
ITERATIVE PROCESSES
Refinement Loop – Cyclical process of prompt testing and improvement based on results
Iterative Co-design – Collaborative refinement through conversation rather than single-shot prompting
DESIGN THINKING
Functional Requirements – Specifying what something needs to do rather than just what it should say
Component Thinking – Breaking complex requests into discrete functional parts
User Flow Specification – Describing the journey through an experience from start to finish
State Management Consideration – Thinking about what information needs to persist, change, or be tracked
Concrete Examples – Providing specific instances to clarify abstract requirements
ORGANIC DEVELOPMENT
Behavioural Emergence – Letting the model shape details organically within your framework
ANTI-PATTERNS
Noise Introduction – Adding unnecessary details that distort results or dilute signal density
Last updated: January 2026