AI/LLM code generation refers to using large language models to assist developers by automatically generating, completing, refactoring, explaining, and debugging code through interaction modes including inline autocomplete, conversational chat, slash commands, and autonomous agents. Tools like GitHub Copilot, Cursor, Claude Code, and Windsurf integrate into development environments (VS Code, JetBrains IDEs, Eclipse, Xcode, or the terminal), powered by models such as Claude Opus 5, GPT-5.5, and Gemini 3.1 Pro. The effectiveness of these tools depends heavily on context management: the AI must understand your codebase through workspace indexing, Copilot Spaces, open files, and explicit references, and on crafting specific, unambiguous prompts. Modern coding agents now operate autonomously, creating branches, running tests, and opening pull requests with minimal intervention. The field keeps consolidating around a handful of serious players, with independent 2026 developer surveys now showing Anthropic's terminal-first Claude Code as the fastest-growing and most-used tool barely a year after launch.
What This Cheat Sheet Covers
This topic spans 20 focused tables and 184 indexed concepts, 95 flashcards. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.
A jump-to index of every table row in this cheat sheet.
An interactive map of every table and concept in this topic.
Table 1: Core Interaction Modes
The way you talk to an AI coding tool shapes what it can do for you. These modes span the full spectrum from passive to autonomous: ghost-text autocomplete that finishes the line you're typing, a chat sidebar for back-and-forth discussion, inline edits at the cursor, and full agent modes that plan and execute multi-step work on their own. Knowing which mode fits the task is the single biggest lever on productivity.
| Mode | Example | Description | |
|---|---|---|---|
Type function validate → ghost text appears | • Real-time ghost text suggestions that appear as you type • accept with Tab, dismiss with Esc• fastest path for single-line completions. | ||
Open chat panel, ask "explain this function" | Conversational sidebar for multi-turn discussions, architecture questions, and complex problem-solving without leaving the IDE. | ||
Ctrl+I / Cmd+I directly in editor | Contextual overlay at cursor position for quick edits, refactors, or explanations without switching to the chat panel. | ||
Assign task → agent plans and executes autonomously | • Autonomous coding that performs multi-step tasks across files, runs terminal commands, iterates on errors, and self-corrects • GA in VS Code and JetBrains (March 2026). | ||
Assign GitHub issue → agent opens PR asynchronously | Background autonomous worker that analyzes the issue, creates a branch, writes code, runs tests, and opens a pull request for review. | ||
gh copilot suggest "find large files" | • Command-line agent with Plan mode, Autopilot mode, parallel subagents, and repository memory • GA February 2026. | ||
Cmd+K → describe change → AI modifies selection | Targeted inline edits (Cursor) for transforming selected code based on natural language instructions. | ||
Add files to working set → describe feature (Cursor) | Multi-file generation (Cursor) that creates or modifies multiple files simultaneously with side-by-side diff review. | ||
Research mode: "Map auth handling across repo" | • Three agent operating modes available from the GitHub Agents tab (April 2026) • Research gathers info without PR; Plan drafts implementation for review; Code writes on a branch. |