AI Concepts, Tools and Agents — A Practical Map
AI is a broad category, and most confusion around it comes from overlapping words: tools, agents, assistants, chatbots, models, coding plans. This guide is the map — it explains how the pieces fit together and links out to the specific guides on this site.
The three layers that matter
When people say "AI" they are usually pointing at one of three layers:
- Models — the engines (GLM, Claude, Gemini, DeepSeek). You rarely touch these directly; they power everything else.
- Tools — applications built on models: chat assistants, image generators, writing helpers, Excel add-ins. You use these directly, one task at a time.
- Agents — systems that take a goal, plan steps, call tools and act with some autonomy instead of answering one question.
Most "which AI should I use" questions are really "which layer am I asking about". A chatbot answers; an agent does. A model is capability; a tool is packaging.
The comparisons people actually search for
- AI agent vs chatbot vs assistant — the difference is autonomy and tool use, not intelligence. scripted replies < single-turn answers < multi-step autonomous work.
- AI tools vs AI models — you buy access to a tool; the model underneath can be swapped. Tool quality (context handling, exports, integrations) often matters more than the model brand.
- AI coding tools compared — editors and agents (Cursor, Claude Code, Codex CLI and friends) now differ more in workflow than in raw model quality. See our coding plan comparison for the subscription side.
Where to go deeper on this site
- DeepSeek guides — using DeepSeek models, API and the Harness desktop app, start to finish.
- Coding Plan comparison — every AI coding subscription compared: prices, quotas, tool compatibility.
- Markdown guides — writing and formatting documentation, which is most of the work AI tools help with.
What this section covers next
This AI section is being built out along four lines: core concepts (agents, tools, models — and how they differ), tool round-ups that stay updated, AI coding workflows, and practical agent use cases. Each guide is written to answer a specific question with a clear recommendation, not to summarize headlines.