AI – Artificial Intelligence
Technology that allows computers to do tasks that normally require human intelligence.
Example: Understanding questions, recognizing images, and writing code.
LLM – Large Language Model
An AI model trained on large amounts of text and code. It can understand and generate human language.
Examples: GPT, Claude, Gemini.
Token
A small piece of text that an LLM processes when understanding or generating language. A token can be a complete word, part of a word or a punctuation mark.
Example: The sentence "Hello, how are you?" is divided into smaller pieces called tokens.
Context Window
The maximum amount of information (measured in tokens) that an LLM can process and use at one time when generating a response.
Example: If an LLM has a context window of 100,000 tokens, it can process up to 100,000 tokens of input and output combined, subject to the model's limits.
LLM Reasoning (Thinking)
The ability of an LLM to analyze information, break complex problems into smaller steps and work toward a solution instead of generating an immediate answer.
Example: An AI model analyzes an Oracle SQL query, identifies performance problems, evaluates possible solutions and suggests an optimized query.
AI Application
A software application that uses AI models to provide useful features and perform tasks.
Examples: ChatGPT (answering questions), Claude (writing and coding), Gemini (understanding text and images), Cursor (AI-assisted code development).
Context Engineering
The process of preparing, organizing and providing relevant information to an AI model so it can generate more accurate and useful answers.
Example: Before asking an AI coding agent to develop a new feature, you provide it with project structure, database schema, coding standards, existing source code and business requirements.
Key idea: Give AI the right information and context so it can understand the project and perform tasks correctly.
AI-assisted development – AI as a coding assistant
A software development approach where AI helps developers write, understand, debug and optimize code, while the developer remains responsible for executing, testing and managing the development process.
Example: A developer provides an Oracle SQL query to ChatGPT and asks it to optimize the query. ChatGPT suggests an improved version, but the developer copies, executes and tests it manually.
Key idea: AI assists the developer, but the developer controls the development workflow.
Vibe Coding
A software development approach where developers use natural language prompts to describe what they want, and AI generates the code and helps build the application, with less manual coding.
Example: You ask ChatGPT or Cursor to create a simple Java application for managing bank customers. AI generates the code, creates files and helps fix errors based on your feedback.
Key idea: Describe what you want to build in natural language and let AI handle much of the coding.
AI Agent
An AI system that can understand a goal, plan what to do, use tools and perform tasks on its own.
Example: Finding a bug, analyzing code, fixing a file and running tests.
Agentic Loop
A repeated cycle in which an AI Agent observes information, thinks about what to do, performs an action using tools, evaluates the result and repeats the process until the task is completed.
Example: An AI coding agent finds a bug, analyzes it, modifies the code, runs tests and checks the results. If the tests fail, it tries again.
Typical cycle: Observe → Think → Act → Evaluate → Repeat
AI Coding Agent
An AI-powered tool that can understand software development tasks, inspect project files, write and modify code, execute commands and run tests using available tools.
Example: You ask Claude Code to optimize an Oracle SQL query. It inspects the project, analyzes the query, modifies the code, runs tests and reports the results.
Key idea: An AI coding agent does more than generate code; it can perform development tasks.
AGENTS.md – AI Agent Instructions File
A Markdown file (AGENTS.md) that provides coding agents with project-specific instructions, rules and context. It helps AI coding agents understand how to work with a software project and follow its development standards.
Location: It is typically placed in the project root directory, but can also be placed in specific subdirectories to provide instructions for particular modules or components.
Example: Creating an AGENTS.md file that tells an AI coding agent to use Java 17, follow specific coding standards, use Oracle 11g, avoid WITH clauses in SQL queries and run tests after modifying code.
YOLO (You Only Live Once)
A mode in AI coding agents where the agent executes actions automatically without asking for human approval before each action.
It allows the AI agent to modify files, execute terminal commands, install packages and run tests with minimal human intervention.
Example: In normal mode, Claude Code asks for permission before executing a terminal command. In YOLO mode, it executes the command automatically.
Ralph Loop
A technique where an AI coding agent repeatedly tries to complete a task. It writes code, checks the result, fixes errors and tries again until the task is completed.
Example: You ask AI to create a PL/SQL procedure. It writes the code, runs tests, finds errors, fixes them and repeats the process.
Typical cycle: Write Code → Test → Fix Errors → Repeat
Key idea: Let AI keep trying until the task is done, instead of manually asking it to try again after every failure.
Autonomous AI development – Multiple agents
A software development approach where multiple specialized AI agents independently perform different tasks and collaborate to complete a larger development project with minimal human intervention.
Example: You ask an AI system to develop a new online banking payment feature. One agent analyzes requirements, another designs the database, another writes Java and PL/SQL code and another tests the implementation. An Orchestrator Agent coordinates their work and combines the results.
Key idea: Multiple AI agents work together like an autonomous software development team, while the developer supervises the overall process.
Orchestrator Agent
An AI agent responsible for coordinating multiple specialized agents, assigning tasks, managing dependencies and combining their results.
Example: In a banking payment development project, the Orchestrator Agent assigns database development to one agent, Java development to another and testing to a third agent.
Key idea: The Orchestrator Agent manages collaboration between multiple AI agents.
