SOLID STATE PRESS
← Back to catalog
AI Agents: When Chatbots Start Doing Things cover
Coming soon
Coming soon to Amazon
This title is in our publishing queue.
Browse available titles
Artificial Intelligence

AI Agents: When Chatbots Start Doing Things

Tool Use, ReAct Loops, and the Leap from Answering to Acting — A TLDR Primer

Your chatbot answers questions. An agent books the flight, writes the code, and fixes its own mistakes along the way — and if you're trying to understand what changed, the explanations online swing between hype and jargon. This guide is for the student, parent, or curious professional who wants a straight answer to what is an ai agent for beginners without wading through a computer science course.

This TLDR primer walks through the shift from chatbot to agent: what an LLM (large language model) needs beyond just generating text — tools it can call, memory it can draw on, and a loop that lets it act, check the result, and try again. You'll see the react loop ai agents explained step by step with a worked trace, so the idea of 'reasoning then acting' stops being an abstraction and becomes something you can trace on paper.

From there, the book covers the real engineering (retrieval, multi-agent setups, short- versus long-term memory), the ways agents fail in production (hallucination, infinite loops, prompt injection attacks), and where these systems are actually deployed today — coding assistants, browser agents, research tools. It closes with the honest version of an ai agents jobs safety guide: what's genuinely at stake for work and accountability as software starts having goals of its own, without overselling or fear-mongering.

No padding, no recycled blog-post fluff — just what you need to walk into a class, an interview, or a dinner-table argument about AI and sound like you know what you're talking about. Start reading and get oriented in one sitting.

What you'll learn
  • Explain the difference between a chatbot and an AI agent in terms of tools, memory, and autonomy.
  • Describe the ReAct loop (reason, act, observe) and how tool-calling APIs let an LLM take real actions.
  • Identify the main failure modes of agents — hallucination, loop-getting-stuck, prompt injection — and basic mitigations.
  • Recognize concrete agent examples (coding agents, browser agents, research agents) and what they can and can't do today.
  • Reason about the safety, economic, and ethical stakes of giving software goals instead of instructions.
What's inside
  1. 1. From Chatbot to Agent: What Actually Changed
    Defines an AI agent by contrasting it with a plain chatbot, and introduces the core ingredients: an LLM, tools, memory, and a goal loop.
  2. 2. The ReAct Loop: How an Agent Thinks and Acts
    Walks through the reason-act-observe cycle with a concrete worked trace, showing how an LLM decides which tool to call and what to do with the result.
  3. 3. Tools, Memory, and the Scaffolding Around the Model
    Explains the engineering that turns a raw LLM into a working agent: tool APIs, retrieval, short- and long-term memory, and multi-agent setups.
  4. 4. Where Agents Break: Hallucination, Loops, and Prompt Injection
    Catalogs the main failure modes with real examples, and covers the mitigations engineers actually use in production.
  5. 5. Agents in the Wild: Coding, Browsing, Research
    Surveys real deployed agents — coding assistants like Devin and Claude Code, browser agents, deep-research tools — and what they can and can't do as of the mid-2020s.
  6. 6. Stakes: Jobs, Safety, and Software That Has Goals
    Frames the near-term economic and safety questions agents raise, including alignment, accountability, and what a student entering this field should watch for.
Published by Solid State Press
AI Agents: When Chatbots Start Doing Things cover
TLDR STUDY GUIDES

AI Agents: When Chatbots Start Doing Things

Tool Use, ReAct Loops, and the Leap from Answering to Acting — A TLDR Primer
Solid State Press

AI Agents: When Chatbots Start Doing Things

Tool Use, ReAct Loops, and the Leap from Answering to Acting — A TLDR Primer

Copyright © 2026 Solid State Press.

Published by Solid State Press.

All rights reserved. No part of this book may be reproduced or transmitted in any form without prior written permission, except for brief quotations in critical reviews and educational use.

Part of the TLDR Study Guides series.

Contents

  1. 1 From Chatbot to Agent: What Actually Changed
  2. 2 The ReAct Loop: How an Agent Thinks and Acts
  3. 3 Tools, Memory, and the Scaffolding Around the Model
  4. 4 Where Agents Break: Hallucination, Loops, and Prompt Injection
  5. 5 Agents in the Wild: Coding, Browsing, Research
  6. 6 Stakes: Jobs, Safety, and Software That Has Goals
Chapter 1

From Chatbot to Agent: What Actually Changed

Ask ChatGPT "What's the capital of France?" and it says "Paris." Ask it "Book me a flight to Paris" and it can only tell you how you might do that — it can't actually open a browser, check prices, or charge your card. That gap is the entire subject of this book.

A large language model (LLM) is a program trained on huge amounts of text to predict, one piece at a time, what word is likely to come next given everything written so far. That's it — underneath the impressive conversation is a very sophisticated autocomplete. A chatbot is an LLM wrapped in a simple interface: you type text in, it predicts and returns text out. The chatbot has no hands. It can describe the world, but it cannot touch it.

An agent is an LLM wired up with three additional things: tools (ways to actually do something outside the conversation — search the web, run code, send an email, edit a file), memory (a way to keep track of what's happened so far across many steps, not just within one reply), and a goal loop (a process that keeps the model working toward an objective across multiple actions, rather than stopping after one response). Give the model those three ingredients, and "book me a flight to Paris" stops being a request for advice and becomes a task the software can actually attempt: search flights, compare prices, fill out a form, report back.

The key shift is from autonomy — the capacity to take actions and make decisions without a human approving each individual step — to goal-directed behavior, meaning the system is oriented around accomplishing an outcome, not just producing a single correct-sounding sentence. A chatbot's job ends the moment it finishes its reply. An agent's job ends when the goal is done — which might take one step or fifty, and which the agent itself has to figure out.

About This Book

If you're a high school student curious about what is an AI agent for beginners, a college freshman in an intro CS or AI course, a developer who wants a ChatGPT agents vs chatbots guide before building something, or a parent trying to understand what your kid means when they say an AI "did their homework," this book is for you.

This is a concise, plain-English answer to how do AI agents work, explained without hype or heavy math. You'll get the ReAct loop AI agents explained step by step, an AI agent tools memory primer covering how models call APIs and remember context, a section with prompt injection explained simply through real examples of agents getting tricked, and an AI agents jobs safety guide look at what autonomous software means for work and risk. A concise overview with no filler, built to get you oriented fast.

Read it straight through first. Then revisit the worked examples, and try the short problem set at the end to check what actually stuck before your exam, interview, or next project.

Keep reading

You've read the first half of Chapter 1. The complete book covers 6 chapters — readable in one sitting.

Coming soon to Amazon