AI Glossary
Complete glossary of AI terms explained simply.
A
Agent Loop
The control flow that lets an LLM act as an agent — repeatedly thinking, calling a tool, observing the result, and deciding the next move.
Agent Orchestration
The discipline of wiring multiple AI agents into a workflow — picking patterns (router, sequential, parallel, hierarchical, swarm) that survive production.
Agentic AI
Agentic AI is the marketing term for systems where LLMs plan multi-step actions, call tools, and pursue goals — what engineers call AI agents.
Agentic Workflow
A multi-step task executed by one or more AI agents that plan, call tools, and adapt — the production shape of agent-based automation.
AI Agent
An autonomous program powered by an LLM that can plan, make decisions, use tools, and complete tasks without constant human oversight.
Attention Mechanism
The core operation in Transformers that lets each token weigh the importance of every other token when computing its representation.
C
Chain-of-Thought (CoT)
A prompting technique that asks the model to produce intermediate reasoning steps before the final answer, improving accuracy on multi-step problems.
Context Window
The maximum number of tokens an LLM can process in a single request, including both the input prompt and the generated output.
F
Few-Shot Learning
A prompting technique where you include 2 to 10 input-output examples in the prompt to steer an LLM toward a specific format, style, or task.
Fine-tuning
The process of further training a pre-trained model on a specialized dataset to improve its performance in a specific domain.
Function Calling
An LLM capability that lets the model produce structured JSON describing a function to call, so your application can execute it and return results.
K
Knowledge Cutoff
The date after which an LLM has no training data. Events, releases, and facts from after the cutoff are unknown to the model unless retrieved at runtime.
Knowledge Distillation
Training a smaller student model to mimic a larger teacher model, producing a faster cheaper model with most of the teacher's capability.
KV Cache (KV Cache)
A memory of past Key and Value attention tensors that lets a Transformer skip recomputing them on every new token during inference.
L
Large Language Model (LLM)
A neural network trained on vast amounts of text data that can generate, understand, and analyze natural language text.
Latency
The delay between sending a request to an LLM and receiving output. In LLM serving, latency splits into time-to-first-token and tokens-per-second.
Low-Rank Adaptation (LoRA)
A parameter-efficient fine-tuning method that trains small low-rank matrices alongside frozen base model weights instead of updating the full model.
M
Model Context Protocol (MCP)
An open standard by Anthropic for connecting AI models to external data sources and tools.
Multi-Agent System (MAS)
Two or more AI agents coordinating on a task — usually through role specialization, message passing, or a coordinator agent that orchestrates the rest.
Multimodal Model
An AI model that natively processes multiple input types — text, images, audio, video — in a single unified model rather than chaining separate specialists.
P
Prompt Engineering
The art and science of crafting instructions for AI models to get optimal results.
Prompt Injection
An attack where adversarial text in user input or retrieved data overrides an LLM's instructions, causing it to leak data or run unintended actions.
R
ReAct (ReAct)
The Reasoning + Acting pattern where an LLM alternates between thinking out loud and calling tools, the foundation of most modern AI agents.
Reasoning Model
An LLM trained or prompted to produce long internal chains of thought before answering, trading inference time for accuracy on hard problems.
Reinforcement Learning from Human Feedback (RLHF)
A training method that aligns LLM behavior with human preferences using a reward model trained on human-ranked outputs.
Retrieval-Augmented Generation (RAG)
A technique that improves LLM responses by adding relevant information from external sources before generating an answer.
S
Subagent
A specialized AI agent spawned by a parent agent to handle a focused task in its own isolated context, then return only the result.
System Prompt
A high-priority instruction passed to an LLM separately from user messages that sets its role, tone, constraints, and behavior for the conversation.
T
Temperature
A sampling parameter that controls randomness in LLM output. Lower values produce focused, deterministic text; higher values produce diverse, creative text.
Token
The smallest unit of text processed by a language model. Roughly 1 token equals 4 characters in English or about 3/4 of a word.
Tokenization
The process of splitting text into tokens — the numeric units an LLM actually reads, writes, and bills you for.
Tool Use
An LLM's ability to call external functions, APIs, or services to gather information or take action, the mechanism that turns a model into an agent.
Transformer
A neural network architecture using self-attention that became the foundation of modern LLMs after Google's 2017 'Attention Is All You Need' paper.