AI Glossary
A practical guide to AI terminology β from foundational concepts to production-ready frameworks. Each term includes real-world examples from our hands-on experience building AI agent systems.
Core Concepts
Agentic AI
What is Agentic AI?
A paradigm where AI doesn’t just passively answer questions but proactively plans, executes, β¦
beginnerAI Agent
What is AI Agent?
An AI program that can autonomously perceive its environment, make decisions, and take actions. Unlike β¦
beginnerAI Alignment
What is AI Alignment?
The research field of ensuring AI behavior aligns with human intentions and values. Misaligned AI β¦
intermediateAttention Mechanism
What is Attention Mechanism?
Allows models to ‘pay attention’ to other relevant words in the input when β¦
advancedChain-of-Thought (CoT)
What is Chain-of-Thought (CoT)?
A technique that makes LLMs reason step-by-step instead of jumping to answers. Like β¦
intermediateConstitutional AI
What is Constitutional AI?
An AI alignment method by Anthropic that uses a set of explicit principles (a β¦
advancedContext Window
What is Context Window?
The maximum number of tokens an LLM can process at once. Claude’s context window is 200K β¦
beginnerEmbedding
What is Embedding?
A technique that converts text, images, etc. into numerical vectors. Distance between vectors β¦
intermediateEmbodied AI
What is Embodied AI?
Giving AI a ‘body’ to perceive and act in the physical world. In 2026, VLA models let β¦
intermediateFine-tuning
What is Fine-tuning?
Training a pre-trained LLM with domain-specific data to specialize it for particular tasks. β¦
intermediateFoundation Model
What is Foundation Model?
Large AI models pre-trained on massive datasets that can be adapted to various downstream β¦
beginnerHallucination
What is Hallucination?
When an LLM confidently generates content that looks correct but is actually wrong. Like a β¦
beginnerInference
What is Inference?
The process of using a trained model to process new inputs and produce outputs. Every time you chat β¦
beginnerLLM (Large Language Model)
What is LLM (Large Language Model)?
An AI model trained on massive text data that can understand and generate human β¦
beginnerMulti-Agent System
What is Multi-Agent System?
An architecture where multiple AI Agents collaborate, each handling different specialties. β¦
intermediateParameters
What is Parameters?
Learnable values inside a model that determine its ‘knowledge’. GPT-4 reportedly has 1.8 β¦
intermediatePrompt Engineering
What is Prompt Engineering?
Designing input instructions that get the best results from LLMs. Good prompts can make the β¦
beginnerRAG (Retrieval-Augmented Generation)
What is RAG (Retrieval-Augmented Generation)?
A technique that lets an LLM search for relevant data before answering, β¦
intermediateRLHF (Reinforcement Learning from Human Feedback)
What is RLHF (Reinforcement Learning from Human Feedback)?
A method where humans rate AI responses, then use those β¦
advancedTokenization
What is Tokenization?
The process of breaking text into small units (tokens) that models can process. English: ~1 word β β¦
beginnerTool Use / Function Calling
What is Tool Use / Function Calling?
Allowing LLMs to call external tools (search engines, databases, APIs) to complete β¦
beginnerTraining
What is Training?
The process of having a model learn patterns and knowledge from data. Pre-training an LLM requires β¦
intermediateTransformer
What is Transformer?
A deep learning architecture proposed by Google in 2017 that uses ‘attention mechanism’ β¦
intermediateVector Database
What is Vector Database?
A database specialized for storing and searching embedding vectors. Traditional databases β¦
intermediateWeights
What is Weights?
The specific values of model parameters, continuously adjusted during training. Open-source models β¦
intermediateFrameworks & Tools
Agent Executor
What is Agent Executor?
A framework component that manages Agent lifecycles β starting, executing tasks, handling β¦
intermediateAPI Gateway
What is API Gateway?
A unified entry point managing all API requests, handling authentication, rate limiting, routing, β¦
intermediateClaude Code
What is Claude Code?
Anthropic’s official CLI tool that lets Claude work directly in your terminal β β¦
beginnerExtended Thinking
What is Extended Thinking?
A feature that lets LLMs perform deep internal reasoning before answering. Claude’s β¦
intermediateFunction Calling
What is Function Calling?
An LLM API feature that lets models output structured function call requests instead of plain β¦
intermediateHooks (Claude Code)
What is Hooks (Claude Code)?
Claude Code’s event interception system that auto-runs custom scripts before/after β¦
intermediateKnowledge Base
What is Knowledge Base?
An organized information repository for Agents to query and reference. Can be document β¦
beginnerMCP (Model Context Protocol)
What is MCP (Model Context Protocol)?
An open protocol by Anthropic that standardizes how AI Agents connect to external β¦
beginnerMemory System
What is Memory System?
A mechanism that lets Agents retain information across conversations. Divided into short-term β¦
intermediatePrompt Caching
What is Prompt Caching?
Caching repeated prompt prefixes to avoid reprocessing the same system prompt and context every β¦
intermediateSDK (Software Development Kit)
What is SDK (Software Development Kit)?
A library and toolset that lets developers quickly integrate specific services. β¦
beginnerSkills (Claude Code)
What is Skills (Claude Code)?
Reusable instruction templates for Claude Code, defined in Markdown with workflows and β¦
intermediateStreaming
What is Streaming?
LLM returns results as they’re generated, rather than waiting until completion. Better UX β¦
intermediateStructured Output
What is Structured Output?
Making LLMs output data in a specified JSON Schema instead of free-form text. Critical for β¦
intermediateInfrastructure
API Rate Limiting
What is API Rate Limiting?
Limiting the number of API requests a client can make within a specific timeframe. Protects β¦
intermediateCI/CD
What is CI/CD?
Automated code testing and deployment workflows. CI runs tests automatically after every code commit, CD β¦
intermediateCron Job
What is Cron Job?
Linux’s scheduled task mechanism that automatically runs scripts at specified times. The unsung β¦
beginnerDocker
What is Docker?
Technology that packages applications with all dependencies into ‘containers’, ensuring β¦
beginnerEdge AI
What is Edge AI?
Running AI inference on local devices (phones, IoT) instead of the cloud. Advantages: low latency, β¦
intermediateFederated Learning
What is Federated Learning?
A technique where multiple parties collaboratively train a model without sharing raw data. β¦
advancedGPU (Graphics Processing Unit)
What is GPU (Graphics Processing Unit)?
Core hardware for AI model training and inference. NVIDIA A100/H100 are β¦
beginnerModel Serving
What is Model Serving?
Deploying trained AI models as services that accept API requests. Many options: run locally with β¦
intermediateREST API
What is REST API?
The most common Web API design style, using HTTP methods (GET/POST/PUT/DELETE) to operate on β¦
beginnerWebSocket
What is WebSocket?
A protocol enabling bidirectional real-time communication between server and client. Unlike HTTP β¦
intermediateSecurity
Access Control
What is Access Control?
Security mechanisms controlling who can access what resources. Especially critical in AI Agent β¦
beginnerAdversarial Attack
What is Adversarial Attack?
Making small but carefully crafted modifications to AI model inputs, causing completely β¦
advancedAPI Key Security
What is API Key Security?
Practices for protecting API keys from leakage and abuse. Use .env files, never hardcode, β¦
beginnerData Poisoning
What is Data Poisoning?
Injecting malicious data into model training data, making the model learn wrong patterns or β¦
intermediateJailbreak
What is Jailbreak?
Techniques to bypass AI safety restrictions, making models output content they should refuse. Unlike β¦
beginnerOWASP Top 10 for LLM
What is OWASP Top 10 for LLM?
A list of the top 10 security risks for LLM applications, maintained by OWASP. Includes β¦
intermediatePrompt Injection
What is Prompt Injection?
Attackers craft malicious inputs to make AI ignore its original instructions and perform β¦
beginnerRed Teaming
What is Red Teaming?
Testing AI system security by simulating attacker perspectives. Professional red teams try prompt β¦
intermediateSandbox
What is Sandbox?
A security mechanism that restricts programs to operate within specific boundaries. In AI Agent β¦
beginnerSupply Chain Attack
What is Supply Chain Attack?
Compromising target systems through third-party packages, plugins, or dependencies. In the β¦
intermediateQuantitative Trading
Backtesting
What is Backtesting?
Testing trading strategies against historical market data. The key is avoiding overfitting β a β¦
beginnerBollinger Bands
What is Bollinger Bands?
A channel formed by a middle line (SMA) and upper/lower standard deviation bands. Price β¦
beginnerEMA (Exponential Moving Average)
What is EMA (Exponential Moving Average)?
A weighted moving average that gives more weight to recent data. We use 4 EMAs β¦
beginnerLive Trading
What is Live Trading?
Trading with real money. The biggest differences from paper trading are psychological pressure and β¦
beginnerMACD (Moving Average Convergence Divergence)
What is MACD (Moving Average Convergence Divergence)?
Uses the difference between two EMAs of different periods to judge β¦
intermediateMarket Regime
What is Market Regime?
The current market state β trending, ranging, or high volatility. Different regimes need β¦
advancedPaper Trading
What is Paper Trading?
Simulating trades with virtual funds on real market data, testing strategies without real risk. A β¦
beginnerPosition Sizing
What is Position Sizing?
Methods for determining how much capital to allocate per trade. No matter how good the β¦
intermediateQuantitative Trading
What is Quantitative Trading?
Using mathematical models and code to automate trading decisions, replacing human β¦
beginnerRisk Management
What is Risk Management?
Systematic methods for controlling trading losses. Our risk control includes: max 2% loss per β¦
beginnerRSI (Relative Strength Index)
What is RSI (Relative Strength Index)?
An indicator measuring overbought/oversold conditions, ranging 0-100. RSI > 70 β¦
beginnerStop Loss
What is Stop Loss?
A preset exit price that automatically closes a position when losses reach a threshold. Set β¦
beginnerStrategy Engine
What is Strategy Engine?
The core module that scans market data, calculates indicators, and generates trading signals. β¦
intermediateTake Profit
What is Take Profit?
A preset profit-taking exit price. We use staged take profits: TP1 (1.5:1 R:R) closes 25%, TP2 β¦
beginnerWalk-Forward Optimization (WFO)
What is Walk-Forward Optimization (WFO)?
A validation method to prevent backtesting overfitting. Data is split into β¦
advanced