What Are Tokens and How Do They Affect Language Models? 🪙
What tokens are in language models, how they affect the context window, and why they matter for working effectively with LLMs.
← Paths
Guided learning path
Understand AI
From the units a model processes to a complete agent system—a path about mechanisms, not just tools.
- Core
- 10
- Deep dive
- 6
Who this path is for
For readers who want to understand what happens inside AI systems before moving into building or automation.
What you will know
By the end, you can explain how models represent information, learn, use external knowledge and connect to tools and agents.
How a model sees information
Tokens, variation, meaning and differences between model families.
This chapter starts withStart with the smallest unit a model processes. Without tokens, context and model limits are hard to understand.
Why nowOnce you know what enters the model, learn why the same input can produce different answers.
How Does Temperature Affect AI? 🌡️
How temperature affects the responses of language models, and how you can simulate different creativity levels through prompt phrasing.
Why nowTemperature explains variation, while embeddings explain how a model represents and compares meaning.
What Is Embedding? How Spotify Knows Exactly Which Song You'll Love Next 🎤
What is Embedding, how does it work behind the scenes, and how does this technique power the recommendation engines of streaming platforms like Netflix and Spotify, and search engines like Google?
Why go deeper hereAfter meaning representation in language, comparing diffusion models shows that not every generative model creates in the same way.
What's the Difference Between GPT's Image Generation Model and a Diffusion Model?
A comparison between diffusion models and GPT-4o: what's the difference between AI image generation methods? A concise guide to understanding the advantages, disadvantages, and best use cases for each approach to AI image generation.
How models learn
Examples, patterns, rewards and efficient model adaptation.
This chapter starts withAfter representing information, move to the ways systems learn patterns from data.
What Is the Difference Between Supervised and Unsupervised Learning in Machine Learning 🤖📊
A clear explanation of the difference between supervised and unsupervised learning, with real-world examples from organizations and AI.
Why nowNot all learning relies on labeled examples. Here the focus shifts to actions, rewards and trial and error.
What Is Reinforcement Learning — The Way to Train an AI 🤖🐶
What is reinforcement learning, how an agent learns from rewards and mistakes, and how it relates to training real-world AI systems.
Why nowTraining builds capability. Zero-shot and few-shot techniques show how to guide it at use time.
How Does AI Learn? Zero-shot & Few-shot Learning
What are Zero-shot and Few-shot Learning, how AI models perform tasks with minimal examples, and how to use them in prompts.
Why go deeper hereAfter guidance at use time, LoRA shows another route: adapting an existing model without retraining it from scratch.
What Is LoRA? The Technique That Saves Time ⏳ and Money 💰 in Language Model Training
What LoRA is, how it reduces the cost and speeds up the fine-tuning of language models, and when to consider it instead of full model training.
Reasoning, knowledge and trust
Breaking down problems, connecting sources and understanding the risks.
This chapter starts withAfter examples and guidance, learn how request structure helps break a problem into steps.
Chain of Thought (CoT) – The Method Behind Reasoning Models 🔗
What is Chain of Thought, how does it relate to Reasoning models, and when should you ask a model to explain its steps rather than give a short answer.
Why nowReasoning is not enough when information lives outside the model. RAG adds real, current knowledge.
What Is a RAG Chatbot and How Does It Improve Our User Experience? 🛒
What is a RAG Chatbot, how it connects a language model to a knowledge base, and why it improves answers, reliability, and user experience.
Why go deeper hereExternal knowledge improves answers but also opens an attack surface. Now examine the risks it creates and how to reduce them.
2 Security Vulnerabilities in LLMs — And What Do the Avengers Have to Do With It? 🦾
Two common security vulnerabilities in large language models, why they are dangerous, and how to think about safer LLM use in your organization.
Connecting to systems
From answering questions to reading information and taking action.
Why go deeper hereFor a model to become part of a product, it must exchange information with other systems. An API is the basic connection layer.
What Is an API in AI — and Why Does It Matter So Much?
What is an API, and how is it changing the rules of the game in artificial intelligence? A clear guide with examples, tools, and tips — from GPT to AI agents that work just like real employees.
Why nowAn API connects a specific system. MCP defines a shared interface through which agents discover tools, files and information sources.
What Is MCP and How It's Changing the Way We Use AI
Meet MCP – a groundbreaking protocol that enables AI agents to access tools, files, and external systems in a smart, context-aware way, without the need for custom integrations. This isn't just another API – it's a new way of working with artificial intelligence.
From a simple agent to an agentic system
Autonomy, operating instructions and the engineering that holds it together.
Why go deeper hereAfter connecting tools, decide who controls the sequence: a fixed workflow or an agent that makes decisions.
What Is the Difference Between Automation Workflow and Agentic Workflow
A clear explanation of the difference between standard automation and Agentic Workflow, and when to use each approach in your organization.
Why go deeper hereGreater autonomy requires clear operating instructions. A Skill defines an agent's goal, process, boundaries and checks.
What Are Skills in Working with AI Agents
If an AI agent is a new employee, a skill is their instruction sheet: what the goal is, what information matters, how to work, where to stop, and how to verify the result is good.
Why nowHere every part connects into an agent with context, tools, memory, permissions, control and observability.
What Is Agentic Engineering and How Do You Build Autonomous AI Agents?
The difference between a chatbot and an autonomous AI agent, what an agent harness is, which components a real agent must have, and how Claude Code Agents, OpenClaw, and Hermes differ from one another.
You completed the core path. Where next?
Continue with the remaining deep dives or choose another path.
You also completed every deep dive—the entire playlist is done.