Objective: Master key terminology for AI product development.
This glossary defines essential terms used throughout the AI-Powered Products book. Each term includes cross-references to relevant chapters for deeper exploration.
Agentic AI
AI systems capable of autonomous decision-making and action execution, often involving multiple tool use and multi-step task completion.
Chain-of-Thought (CoT)
A prompting technique where the model is encouraged to show intermediate reasoning steps before giving a final answer.
Context Window
The maximum amount of text (measured in tokens) that an LLM can process in a single request, including both input and output.
Eval (Evaluation)
A systematic assessment of AI output quality, often using LLM-as-Judge or human raters to score responses against defined criteria.
Fine-tuning
Training a pre-existing foundation model on domain-specific data to improve performance on targeted tasks.
Foundation Model
Large language models trained on broad data that can be adapted to many downstream tasks through prompting or fine-tuning.
Grounding
The process of connecting AI outputs to reliable external knowledge sources to reduce hallucinations.
Hallucination
When an AI model generates content that appears plausible but is actually incorrect, fabricated, or unrelated to the input.
Hybrid Search
An information retrieval approach combining vector semantic similarity with traditional keyword matching for improved relevance.
Inference
The process of running a trained AI model to generate predictions or outputs from given inputs.
Intent Classification
The task of mapping user input to a discrete category representing the user's intention or goal.
LLM-as-Judge
Using a larger or more capable language model to evaluate and score the outputs of other AI systems.
MCP (Model Context Protocol)
A protocol enabling AI models to connect with external tools and data sources in a standardized way.
Multimodal AI
AI systems capable of processing and generating multiple modalities including text, images, audio, and video.
One-shot Learning
An approach where an AI model performs a task based on a single example provided in the prompt.
PEEK Method
A UX design framework for AI products: Predict, Explain, Execute, Keep in loop. Guides how users maintain control over AI actions.
Prompt Injection
An adversarial attack where malicious instructions are embedded in user inputs to manipulate AI system behavior.
RAG (Retrieval-Augmented Generation)
An architecture combining retrieval systems with language models to ground outputs in external knowledge.
Reasoning Models
AI models that use chain-of-thought reasoning to break down complex problems before generating answers, typically offering higher accuracy at the cost of increased latency.
Temperature
A sampling parameter controlling output randomness. Lower values produce more deterministic outputs; higher values increase creativity and variation.
Token
The basic unit of text that language models process, roughly equivalent to 4 characters or 0.75 words in English.
Tool Use
An AI model's ability to invoke external functions or APIs to accomplish tasks beyond text generation.
Vibe Coding
High-velocity AI-assisted development where human intention is rapidly translated to working code through conversational AI interaction.
Zero-shot Learning
An approach where an AI model performs a task without any examples, relying only on instructions in the prompt.