Module 1: Introduction to Vectors

1. Module Introduction: Vectors - The Foundation of Everything

Overview

This is the first module of the AI Semantics Guide and the starting point for understanding how semantic search works and why RAG works. You won't be writing Python code or implementing embeddings here: you're going to understand what vectors are conceptually and why they're the heart of modern artificial intelligence.

A lot of people talk about "embeddings", "vector databases" and "semantic search" without being clear on what a vector is or why they're used. Over the next capsules you'll build that solid conceptual foundation: from simple 2D vectors to vectors in very high dimensions (1536D, 3072D), and you'll understand why representing meaning as numbers is the key to all of it.

The goal is for you to have conceptual clarity before implementing code. Without this foundation, when you see openai.embeddings.create() in Guide #6 (Embeddings Deep Dive), you'll just be calling an API without understanding what it does under the hood. With this foundation, you'll understand that you're converting text into a high-dimensional vector that captures meaning, and you'll be able to make better design decisions.

This module is designed to be read in about 2-2.5 hours (reading + final exercise). It requires no programming and no advanced math; only spatial visualization and conceptual reasoning. If after this module someone asks you "what is a vector in AI?" or "why do we use vectors to represent text?", you'll be able to answer precisely and without resorting to formulas.

Why this module is different from other courses: Most courses on embeddings or semantic search make two mistakes: (1) they jump straight to code without explaining WHAT vectors are, or (2) they require advanced linear algebra. This module sits in the middle: it gives you the geometric and conceptual intuition without complex formulas, preparing you to implement technically afterward without staying on the surface or overwhelming you with math.

Analogy: The map before the journey

Think of this module as the dictionary of a new language. You can't speak the language without knowing the basic words. In modern AI, "vector" is the most fundamental word: embeddings are vectors, semantic search compares vectors, vector databases store vectors, RAG searches for similar vectors. If you don't know what a vector is, all of those terms are black boxes.

This module gives you the dictionary: what a vector is (basic geometry), how they're visualized (2D, 3D), what happens in high dimensions (100D, 1536D), what basic operations exist (addition, subtraction, scaling), and why AI uses them to represent meaning (words, phrases, documents). With this dictionary, when you read "vector space" or "cosine similarity" in later modules, they won't be abstract terms but concepts you've already visualized.

Context: Where are we in the guide?

This guide has 8 modules across 3 phases. Each phase has a clear objective:

Phase 1 — Vector Fundamentals (Modules 1-4): You are here. In this phase you build the foundation: vectors (Module 1), vector spaces (Module 2), similarity and distance (Module 3), proximity search (Module 4). All conceptual, no code. The goal is that when we talk about "semantic search" in Phase 2, or about "embeddings" in Guide #6, they aren't loose words but concepts already placed on your mental map.
Phase 2 — Semantic Search (Modules 5-6): Keyword search vs semantic search, and the architecture of semantic search systems. Here you apply what you learned in Phase 1: vectors (Module 1) are used to represent documents and queries, similarity (Module 3) is used to find relevant documents, proximity search (Module 4) is used to scale to millions of documents.
Phase 3 — Application to AI Engineering (Modules 7-8): The connection to RAG and real use cases. Here you close the loop: you understand why RAG works (it uses semantic search for retrieval), and you design complete systems applying everything above.

Module 1 is the front door. It doesn't assume you know linear algebra or advanced programming; only that you want to understand the fundamentals before implementing. If you already have some context (for example, you've used ChatGPT or you've heard of "embeddings"), even better; but it isn't required.

Why the 3-phase structure matters: Phase 1 (where you are) gives you the building blocks: vectors, spaces, similarity, search. Without that, Phase 2 (semantic search) would be a list of tools with no foundation. Phase 2 gives you the "how it's used" (comparing keyword vs semantic, designing architectures); without that, Phase 3 (RAG, use cases) would be abstract. Phase 3 connects everything to real applications: what an AI Engineer actually does with vectors and semantic search. Each phase builds on the previous one; that's why this first module is fundamental.

Why this module is the foundation (concrete examples): When you read "cosine similarity" in Module 3, you'll already be clear on what a vector is and how to visualize it; without this module, "cosine" would just be a formula with no intuition. When you read in Module 5 that "semantic search uses vectors to represent meaning", you'll already have seen in this module WHY vectors capture semantic relationships (words close in meaning → vectors close in space). When you read in Module 7 that "RAG uses vector retrieval", you'll already be clear on the full flow: document → vector (embedding) → comparison against the query vector → similar results. Without this module, all of that would be magic; with it, it's clear logic.

Module objective

By the end of this module you'll be able to:

Define precisely what a vector is geometrically (direction, magnitude) and conceptually (numerical representation). In practice: when someone says "embeddings", you'll understand they're high-dimensional vectors; when you see [0.5, -0.3, 0.8], you'll recognize it as a 3D vector and you'll be able to picture it mentally.
Visualize vectors in 2D and 3D and understand how that extends to high dimensions (100D, 1536D, 3072D). In practice: you'll be able to draw 2D/3D vectors on paper representing concepts ("dog", "cat"), and even though you can't draw 1536D, you'll understand that the same ideas of "closeness" and "direction" apply.
Understand basic operations (addition, subtraction, scalar multiplication) conceptually, without complex formulas. In practice: when you read that "embeddings can be added" (e.g. vector("king") - vector("man") + vector("woman") ≈ vector("queen")), you'll understand what that means geometrically and why it works.
Explain why vectors in AI — why representing words, phrases or documents as vectors is the foundation of semantic search and RAG. In practice: you'll be able to justify design decisions ("we use embeddings because they capture semantic relationships better than keyword matching") without resorting to buzzwords.
Draw and analyze vectors conceptually in the final exercise. In practice: you'll complete an exercise where you map 5 concepts to 2D vectors, analyze which ones are near/far, and reflect on how that relates to semantic search.

All of it is conceptual. There's no Python code and no complex linear algebra; there are visualizations, mental diagrams and analysis exercises that reinforce the ideas. If your goal is to implement embeddings or vector databases, that comes in Guides #6 and #7; here we're only laying the foundation so that you don't implement blindly.

Prerequisites

For this module all you need is:

AI Fundamentals Guide completed (or at least Modules 1-2: what AI is, what ML is, what an LLM is). You don't need to know how Transformers work in detail; only that language models exist and that they "learn from data".
Basic intuitive geometry: Understanding that a point on a 2D plane has coordinates (x, y), and that a point in 3D space has coordinates (x, y, z). You don't need to remember the Pythagorean theorem or trigonometry; only basic spatial visualization.
A willingness to draw and visualize: This module is highly visual. You'll have to imagine (or better, draw on paper) vectors, spaces, and relationships. If you prefer learning by reading text alone without mental diagrams, this module will be harder for you.
Optional: Paper and pencil for diagrams. Many concepts stick better when you draw vectors by hand than when you only read about them.

You do not need:

Advanced linear algebra: There are no matrices, no determinants, no formal vector spaces. Everything is explained with intuitive geometry.
Python programming: This module has no code. The examples are conceptual or illustrative pseudocode. Code comes in Guide #6 (Embeddings Deep Dive).
Knowledge of embeddings or vector databases: We start from zero. If you already know these topics, this module will help you solidify the conceptual intuition you might be missing.
Having used AI APIs: You don't need to have called openai.embeddings.create() or to have seen a real embedding. The examples in the text are enough.

If you already have a background in math, physics or engineering, this module will feel very fast (you'll be able to read it in 1-1.5 hours). If you come from non-technical profiles (product, business, design), the module is designed for you: no formulas, with real-world analogies.

If you have X, even better (but not required): If you've already played with ChatGPT or Claude, you'll have the intuition that they "understand meaning" beyond exact words; that helps in capsule 06 (why vectors in AI). If you've already drawn 2D charts or played with visualization tools (Excel, Matplotlib, Tableau), visualizing vectors will come more naturally. If you've already heard of "embeddings" or "vector databases", this module will organize concepts you may have had jumbled together.

If you don't have Y, it doesn't matter: You don't need to know what an "embedding" is (we explain it here and go deeper in Guide #6). You don't need to have ever used a vector database (that's Guide #7). You don't need to know how "cosine similarity" works (that's Module 3). This module only covers basic vectors; everything else comes later.

Module roadmap: the 7 capsules

Here's the map of the 7 capsules, what you'll see in each one and what you'll gain by finishing it.

#	Capsule	What you'll see	What you'll gain	Approx. duration
01	Module introduction	Guide context, module objective, roadmap, limits	Clarity on what this module covers and how it fits into the full guide	15 min
02	What is a vector?	Geometric definition: direction, magnitude, numerical representation	Clear intuition of what a vector is; you'll be able to explain it without formulas	20 min
03	Vectors in 2D and 3D	Visualization, Cartesian coordinates, diagrams	The ability to draw and visualize 2D/3D vectors on paper	25 min
04	Vectors in high dimensions	4D, 100D, 1536D: what changes and what stays the same	Understanding that high dimensionality is an extension of 2D/3D, not magic	20 min
05	Basic operations	Addition, subtraction, scaling (conceptual, no algebra)	Understanding what "adding vectors" means geometrically	20 min
06	Why vectors in AI?	Numerical representation of meaning; words → vectors	The justification for why embeddings work	25 min
07	Capstone exercise	Drawing and analyzing vectors conceptually	Practical application: mapping concepts to vectors, analyzing closeness	30 min

Module total: ~2-2.5 hours (reading + exercise). If you want to go deeper with the additional resources in each capsule, you can add 30-60 minutes more.

How to gauge your own pace: The durations in the table are indicative (medium reading pace, no long breaks). If you read more slowly or want to draw every diagram on paper, it could be 3 hours. If you already have a math background and are only reviewing concepts, it could be 1.5 hours. What matters isn't speed but that by the end you can meet the "Evidence of success" you'll see further down.

Estimated time per capsule (detail): Capsule 01 (this one) usually takes 15-20 minutes if you read calmly. Capsule 02 (what is a vector?) is short but dense; 20-25 minutes. Capsule 03 (2D and 3D) has diagrams; if you draw them on paper, 30 minutes; if you only read, 20 minutes. Capsule 04 (high dimensions) is conceptual; 15-20 minutes. Capsule 05 (operations) is short; 15-20 minutes. Capsule 06 (why in AI) ties everything together; 25-30 minutes. Capsule 07 (the exercise) depends on you: drawing 5 concepts as vectors and analyzing them can take 30-40 minutes. In total, 2-2.5 hours is a good range for a first pass.

Why this capsule order

The capsules are ordered deliberately:

Definition (02) before visualization (03): To visualize vectors you need to know WHAT you're visualizing. If you jump straight to coordinates (x, y, z) without understanding that a vector has direction and magnitude, the numbers make no sense.
2D/3D (03) before high dimensions (04): Geometric intuition comes from spaces you can see. Once you can picture a 3D vector, extending to 100D or 1536D is conceptually "more of the same". Without the 2D/3D intuition, high dimensionality looks like magic.
Visualization (03) before operations (04-05): To understand what "adding two vectors" means, you need to be able to visualize them first. If you read the operations without prior visualization, they stay as abstract rules.
Operations (05) before the AI application (06): In AI, operations like "vector(king) - vector(man) + vector(woman)" are used to manipulate meaning. To understand that, you need to know what vector addition and subtraction are geometrically. Without capsule 05, capsule 06 can feel like "anything goes" with no foundation.
Theory (02-06) before the exercise (07): In the exercise you're going to map real concepts (words, products) to 2D vectors and analyze them. That requires having seen what a vector is, how to visualize it, and why it works in AI. Without capsules 02-06, exercise 07 stays at "drawing dots" with no connection to semantic search or RAG.

If you already have experience with geometry or linear algebra and want to go straight to "why vectors in AI" (capsule 06), you can skip ahead; but on a first pass we recommend reading in order to build the intuition gradually.

Connection to the rest of the guide

What you learn in this module gets used like this in the following ones:

Module 2 (Vector Spaces): When we talk about a "vector space" as a "universe of meaning", you'll already be clear that a vector is a point in that universe with coordinates. That avoids the typical confusion of thinking that "vector space" is something mystical; it's just the set of all possible vectors in N dimensions.
Module 3 (Similarity and Distance): To measure "closeness" between vectors (cosine similarity, Euclidean distance), you need to know what a vector is and how to visualize it. Without this module, "cosine" would just be a formula with no intuition; with it, you'll understand that it measures the "angle" between two vectors (similar directions = similar meaning).
Module 4 (Proximity Search): When we talk about k-Nearest Neighbors (k-NN) or indexes like HNSW, you'll already be clear that we're searching for nearby vectors in space. The operations from this module (addition, subtraction) help you understand how data structures are built to search efficiently.
Modules 5-6 (Semantic Search): When we see that semantic search uses vectors to represent documents and queries, you'll already be clear on WHY: vectors capture semantic relationships (capsule 06) and allow mathematical comparisons (Module 3). Without this module, "embeddings" would be a black box.
Module 7 (RAG): You'll understand that RAG uses vectors in the retrieval step: document → embedding (vector) → comparison against the query vector → top K results. The operations and visualizations from this module underpin that entire flow.
Guides #6, #7, #8 (Technical implementation): When you call openai.embeddings.create() in Guide #6 and get back an array of 1536 numbers, you'll know it's a vector in 1536D that captures the meaning of the text (capsules 04 and 06). When you use ChromaDB or Pinecone in Guide #7, you'll understand they store vectors and search for nearest neighbors (Module 4). When you build RAG in Guide #8, the full vector flow will be transparent to you.

In short: if someone asks you "what is a vector in the context of AI?" or "why do we use vectors to represent text?", after this module you'll be able to answer clearly. And when you read "embedding", "vector database" or "cosine similarity" in later modules, you'll have the conceptual foundation to understand what's being talked about.

Connection summary (reference table):

When in a later module...	You'll use from this module...
Module 2: "vector space", "dimensions"	Definition of a vector, 2D/3D visualization, high dimensionality
Module 3: "cosine similarity", "Euclidean distance"	Vectors as directions, operations (addition, subtraction)
Module 4: "k-NN", "proximity search"	Vectors in spaces, geometric closeness
Module 5-6: "semantic search", "embeddings"	Why vectors represent meaning (capsule 06)
Module 7: "RAG", "vector retrieval"	The full flow: text → vector → comparison → results
Guide #6: `openai.embeddings.create()`	Understanding that it returns a high-dimensional vector (capsule 04)

Limits: what this module does NOT do

So that you don't expect something we're not going to give you, here's what this module does not cover and, where applicable, where it is covered:

It doesn't get into formal linear algebra (vector spaces, bases, linear independence, linear transformations). We keep everything geometric and intuitive. If in the future you want to go deeper into the formal math, there are dedicated linear algebra courses; this module stays at the intuition level.
It doesn't implement code (Python, NumPy, embeddings). That comes in Guide #6 (Embeddings Deep Dive) where you'll see openai.embeddings.create(), and in Guide #7 (Vector Databases) where you'll use ChromaDB or Pinecone. Here it's concepts only.
It doesn't explain how embeddings are trained (Word2Vec, GloVe, BERT, Sentence-BERT). That's Guide #6. Here we only say that "models learn to convert text into vectors that capture meaning"; how they're trained is a separate topic.
It doesn't compare embedding models (OpenAI vs Cohere vs open-source). That's Guide #6 (technical comparison) and Module 6 of this guide (the API ecosystem). Here we only lay the foundation: what a vector is and why it works for representing meaning.
It doesn't go deep into cosine similarity or distances beyond mentioning that they exist. That's Module 3 of this guide. Here we only look at basic operations (addition, subtraction) and leave vector comparison for later.
It doesn't cover vector databases (ChromaDB, Pinecone, Weaviate). That's Guide #7. Here we only explain that "vectors can be stored and searched"; the technical implementation comes later.

This module only covers conceptual vectors. The "how they're generated" (embeddings), "how they're compared" (similarity), "how they're stored" (vector DBs) and "how they're used" (semantic search, RAG) come in later modules and guides.

Where each topic is covered (quick reference):

What this module does NOT cover	Where it is covered
Formal linear algebra (spaces, bases, transformations)	Math courses; here we keep geometric intuition
Python code (NumPy, embeddings)	Guide #6: Embeddings Deep Dive (technical implementation)
How embeddings are trained (Word2Vec, BERT)	Guide #6: Embeddings Deep Dive (conceptual training + code)
Model comparison (OpenAI vs Cohere)	Guide #6 (technical) and Module 6 of this guide (ecosystem)
Cosine similarity, distances (formulas, comparisons)	Module 3: Similarity and Distance (next in this guide)
Vector databases (ChromaDB, Pinecone, code)	Guide #7: Vector Databases Comparison (implementation)

Evidence of success: how you know you're ready

You know you've gotten good value out of the module when:

You can explain in 2-3 sentences what a vector is geometrically (direction + magnitude) and conceptually (numerical representation of a point in N-dimensional space).
You can draw on paper a 2D vector with coordinates (x, y) and a 3D vector with coordinates (x, y, z), and explain what those coordinates mean.
You understand that high dimensionality (100D, 1536D) is an extension of 2D/3D, not something qualitatively different. You can't visualize it directly, but the ideas of "closeness" and "direction" hold.
You can explain what "adding two vectors" means geometrically (displacement, combination of directions) without resorting to formulas.
You can justify why vectors in AI: because they capture semantic relationships (similar words → nearby vectors) and enable mathematical operations (comparing, searching, adding).
You've completed the exercise in capsule 07: at least 5 concepts mapped to 2D vectors, an analysis of which ones are near/far, and a reflection on how that relates to semantic search.

If any of these still isn't clear, go back to the relevant capsule (02 definition, 03 visualization, 04 high dimensions, 05 operations, 06 why in AI, 07 exercise) and review. You don't need to memorize exact coordinates or formulas; what matters is having clear geometric intuition.

Checklist before moving to Module 2: Before continuing with Vector Spaces (Module 2), check that you can do this without looking at the capsules: (1) Draw a 2D vector on paper with coordinates (3, 2) and explain what it means. (2) Explain what a vector in 1536D is conceptually (even if you can't draw it). (3) Explain what adding two vectors means geometrically. (4) Justify in 2 sentences why vectors are used in AI to represent text. (5) Have completed exercise 07 with at least 5 concepts mapped. If you fail (1) or (2), review capsules 02-04. If you fail (3), review capsule 05. If you fail (4), review capsule 06. If you haven't done (5), do it before continuing; it's what closes the loop of the module.

✓ Quick visual checklist before Module 2:

I can explain what a vector is (direction + magnitude + numerical representation)
I can draw 2D and 3D vectors on paper with coordinates
I understand that 1536D is an extension of 2D/3D (not magic)
I can explain what adding vectors means geometrically
I can justify why vectors are used in AI (they capture meaning)
I completed exercise 07 with at least 5 concepts analyzed

If you checked all 6, you're ready for Module 2. If any is missing, review the relevant capsule.

How to use this module

If it's your first time with vectors: Read capsules 02 through 06 in order; then do 07 calmly. Don't skip 07: the exercise forces you to visualize and analyze vectors conceptually, and that locks in the ideas far better than reading alone.
If you already have some math background (for example, analytic geometry or basic physics): You can skim 02 and 03 (you already know what coordinates are), and go deeper into 04 (high dimensions), 05 (conceptual operations without algebra) and 06 (why in AI). Even so, don't skip 07: the exercise of mapping concepts to vectors connects geometry to AI in a way that isn't obvious.
If you're going to do Guide #6 (Embeddings Deep Dive) afterward: This module is the recommended foundation. Having it done beforehand saves confusion when you see 1536D embeddings or when you read about "nearby vectors capture similar meaning". If you don't have time to do the whole module, at least read capsules 02 (what is a vector), 04 (high dimensions) and 06 (why in AI), and do a short version of exercise 07 (3 concepts instead of 5).
If you get stuck on a capsule: Capsule 03 (2D/3D visualization) is the most important one for intuition. If you can't visualize vectors mentally, draw on paper: X axis horizontal, Y axis vertical, Z axis coming out of the paper (or use a pencil to represent it). Capsule 04 (high dimensions) can be abstract; if it's hard, hold on to this idea: "1536D is like 3D but with 1536 axes instead of 3; I can't draw it, but the ideas of closeness and direction work the same". For 06 (why in AI), if you don't see the connection, think: "if two words mean something similar, their vectors should be close; if they're opposites, they should be far apart; that's what models learn". For 07, if you can't think of 5 concepts, start with simple ones: fruits (apple, banana, orange), animals (dog, cat, fish), or emotions (happiness, sadness, anger).

If you get stuck: where to look for help: (1) In capsule 02, if you don't understand "direction + magnitude", think of an arrow: the arrow points in a direction (e.g. northeast) and has a length (magnitude). (2) In capsule 03, if you can't visualize 3D, use physical objects: a glass on your desk has a position (x, y on the table) and a height (z). (3) In capsule 04, if high dimensionality overwhelms you, accept that you can't draw it but that mathematically it works the same as 2D/3D. (4) In capsule 05, if the operations are confusing, visualize vector addition as "walking": vector A takes you from the origin to point A, vector B takes you from A to point B; A + B takes you straight from the origin to B. (5) In 06, if you don't see why vectors capture meaning, read all of capsule 06 and the additional resources (articles about Word2Vec or conceptual embeddings). (6) In 07, if the exercise feels vague, use the complete sample answer (if we include one) as a template.

Each capsule ends with a brief summary and with additional resources to go deeper. The conceptual exercises come with answer guides or examples. Use all of that to check your own thinking and to confirm that you're applying the concepts coherently.

Frequently asked questions

Do I need to know linear algebra for this module?
No. Everything is explained with intuitive geometry and visualization. If you already know linear algebra, even better; but it isn't necessary.

Is this module useful for someone who doesn't want to code?
Yes. It's designed for anyone who wants to understand the fundamentals of semantic search and RAG conceptually. If your role is product, business or design, this module gives you the vocabulary without demanding code.

Can I skip capsule 03 (2D/3D visualization) if I already know geometry?
You can, but it isn't recommended on a first pass. Capsule 03 doesn't just review coordinates; it connects visualization to AI (how to represent concepts as vectors). Without that, capsules 04-06 can feel abstract.

Is exercise 07 mandatory?
To get real value out of the module, yes. Mapping real concepts to vectors and analyzing them is what closes the loop: you go from "reading about vectors" to "using them to think about meaning". If you're really short on time, do a short version (3 concepts instead of 5) rather than skipping it entirely.

What do I do if an additional resource (link) is broken or in another language?
Most resources are in English (papers, technical blogs). If a link is broken, search for the resource title on Google; they're usually kept up to date at another URL. If you prefer material only in your own language, the capsule text has enough content on its own; the additional resources are optional for going deeper.

Can I do the module across several days?
Yes. You can do capsules 01-03 one day (fundamentals), 04-05 another day (operations), and 06-07 another (AI application). Ideally don't leave too much time between 02 (definition) and 07 (exercise), because the exercise applies those definitions; if many weeks pass, review 02-03 before doing 07.

Does this module have a certificate or assessment?
This guide doesn't issue a certificate; the assessment is your own (Evidence of success, Checklist before moving to Module 2, exercise 07). If you're doing the AI Engineering Bootcamp, this module is supporting material; the bootcamp's assessment is separate.

What to expect from each capsule (in detail)

So you know exactly what you take away from each one:

Capsule 02 (What is a vector?): You'll come away with a clear definition: a vector is a geometric entity with direction and magnitude, represented by numerical coordinates. You'll see visual examples (arrows, points) and understand that [3, 2] is a 2D vector. It includes identification exercises (is this a vector? yes/no, why).
Capsule 03 (Vectors in 2D and 3D): You'll learn to visualize vectors on the Cartesian plane (2D) and in 3D space. You'll see how to draw them, how to read coordinates, and how to think of them as "arrows from the origin" or as "points in space". It includes drawing exercises (draw the vector [2, -1] in 2D).
Capsule 04 (Vectors in high dimensions): You'll understand that 4D, 100D, 1536D are mathematical extensions of 2D/3D. You can't draw it, but the ideas of "closeness" (nearby vectors = similar) and "direction" (vectors in opposite directions = opposites) hold. You'll see why embeddings have 1536D or 3072D (more dimensions = more capacity to capture nuance).
Capsule 05 (Basic operations): You'll learn what adding vectors means (combined displacement), subtracting vectors (difference of position), and multiplying by a scalar (changing magnitude). All geometric, no complex algebra. You'll see visual examples and understand that "vector(king) - vector(man)" is subtracting positions in semantic space.
Capsule 06 (Why vectors in AI?): The capsule that ties everything together. You'll understand that vectors capture semantic relationships because models learn to place similar words close together in space and opposite words far apart. You'll see examples (vector("dog") close to vector("cat"), far from vector("car")) and the justification for why this works for semantic search (comparing vectors instead of exact words).
Capsule 07 (Capstone exercise): You'll apply everything above: you'll pick 5 concepts (words, products, emotions), map them to 2D vectors on paper, analyze which are near/far, and reflect on what that means for semantic search. It includes a self-assessment guide and a complete sample answer.

With this you have the full map of the module before starting capsule 02.

Common mistakes when starting this module

Believing that a "vector" is just an array of numbers: No. A vector has a geometric interpretation (direction, magnitude). [3, 2] isn't just "three and two"; it's a point 3 units along X, 2 units along Y, with a direction from the origin and a magnitude of √(3²+2²). You'll see this in detail in capsule 02; don't skip that capsule.
Looking for Python code or NumPy examples: This module is conceptual; there's no code. The examples are visual or illustrative pseudocode. If you're looking to implement embeddings with code, that comes in Guide #6 (Embeddings Deep Dive).
Thinking that "high dimensionality" is impossible to understand: It isn't. You can't draw high dimensions (1536D), but conceptually they work just like 2D/3D: there are points, there are distances, there are directions. You'll see this in capsule 04; don't get frustrated if you can't visualize it, nobody can; the intuition is enough.
Skipping exercise 07: The exercise forces you to apply the concepts to real examples; without it, the definitions are forgotten fast. Even if you're in a hurry, do at least a short version (3 concepts instead of 5).
Thinking that "vectors in AI" are different from "vectors in math": They aren't. They're the same. Embeddings are ordinary vectors in high-dimensional spaces; there's no special magic. You'll see this in capsule 06: the "trick" is that models learn to place similar concepts close together in space, but the vector itself is standard math.

Summary

This is Module 1 of the AI Semantics Guide (Phase 1: Vector Fundamentals).
The goal is to understand what vectors are conceptually (geometry + visualization + operations + why in AI), without code or complex algebra.
There are 7 capsules: introduction (this one), definition, 2D/3D visualization, high dimensions, operations, why in AI, capstone exercise.
What you learn here is the foundation for vector spaces, similarity, search, semantic search, embeddings, vector databases, and RAG in the following modules and in Guides #6-8.
It does not cover formal linear algebra, Python code, embedding training, or vector database implementation; that's covered in later modules or in the technical guides.
You know you're ready when you can explain what a vector is, visualize 2D/3D, understand high dimensionality, explain operations, justify why in AI, and you've completed exercise 07 with at least 5 concepts.
The 3-phase structure (Fundamentals → Semantic Search → Application) builds progressively; this module is the front door.
Before moving to Module 2, check that you meet the "Evidence of success" and the "Checklist before moving to Module 2" in this capsule.

Summary of the full guide (8 modules)

So you have the complete picture before starting:

Modules 1-4 (Phase 1): Vectors, vector spaces, similarity and distance, proximity search. No code.
Modules 5-6 (Phase 2): Keyword vs semantic search, semantic search architecture. How vectors are used in real systems.
Modules 7-8 (Phase 3): The connection to RAG, real use cases, system design. Practical application of everything above.

This module (Module 1) is only the first piece; modules 2-4 complete Phase 1. You don't need to memorize the list of 8 modules; knowing that there are 3 phases (fundamentals → semantic search → application) is enough to keep you from getting lost.

How to use this introduction if you're short on time

If you're very short on time and want to go straight to the content: read at least the "Module objective" section (what you'll be able to do by the end), the "Module roadmap" table (what you'll see in each capsule) and the "Limits" section (what this module doesn't do). With that you have the minimum needed to not expect something we're not going to give you, and to know what you'll gain in each capsule. The rest of this introduction (context, connection to the guide, FAQs, etc.) you can read when you have more time or when a question comes up.

If you have normal time (2-2.5 hours for the full module): read this whole introduction once; it'll help you not lose the thread when you reach capsule 02 and to know why each capsule is where it is. Then continue in order: 02 → 03 → 04 → 05 → 06 → 07.

Next step

When you finish reading this introduction, move on to Capsule 02: What is a vector?. There you're going to build the fundamental definition: what a vector is geometrically (direction + magnitude) and how it's represented numerically with coordinates. With that you'll have the basic building block you'll use throughout the rest of the module and the guide.

How to know if you're ready for capsule 02:

You're clear on what the objective of Module 1 is (understanding vectors conceptually, without code or algebra).
You know what each capsule covers (02 definition, 03 2D/3D visualization, 04 high dimensions, 05 operations, 06 why in AI, 07 exercise).
You understand that modules 1-4 are conceptual fundamentals and that modules 5-8 build on them (semantic search, RAG, use cases).
You know what this module does NOT cover (formal algebra, Python code, embedding training, technical vector databases) and where each thing is covered.

If any of these still isn't clear, review the relevant section of this introduction before continuing.

Additional resources

3Blue1Brown: "Vectors, what even are they?" — A 10-minute video with excellent visualizations of 2D/3D vectors. A perfect complement to capsule 03. In English.
Khan Academy: "Introduction to Vectors" — A series of short videos on basic vectors. Useful if you want to reinforce geometry. In English with subtitles.
Immersive Math: Linear Algebra — A visual interactive book (you can rotate 3D vectors in the browser). Chapter 1 is perfect for this module. In English.
Understanding Word Vectors (blog post) — A conceptual explanation of why vectors represent meaning, with visualizations. Complements capsule 06. In English.
Visualizing High-Dimensional Data (Jay Alammar) — A blog with diagrams about embeddings and high-dimensional vectors. For after you complete the module. In English.

If something doesn't add up

If you don't understand "direction + magnitude": Capsule 02 clears it up with diagrams and analogies (arrows, displacements). A vector is like an arrow: it points in a direction and has a length.
If you can't visualize 3D: Capsule 03 has diagrams. If it's still hard, use physical objects: a glass on your desk has a position (x, y on the table) and a height (z). That's a 3D "vector" conceptually.
If high dimensionality (1536D) overwhelms you: Capsule 04 explains it: you can't draw it, but it works just like 2D/3D. Accept that it exists mathematically and that the ideas of closeness and direction apply; that's enough to understand embeddings.
If exercise 07 feels vague to you: Use the sample answer (if we include one) as a template. Pick 5 concepts you know (fruits, animals, emotions), place them on a 2D plane according to similarity, and analyze which are near/far. There's no single correct answer; what matters is using the concepts coherently.

With this you have everything you need to start capsule 02. When you finish the full module, come back to this introduction if you want to review the map (objectives, roadmap, connection to the guide, limits).

In one sentence: what a vector is (cheat sheet for review): A vector is a geometric entity with direction and magnitude, represented by numerical coordinates (e.g. [3, 2] in 2D, [1, 0, -2] in 3D, [...] in 1536D). In AI, vectors represent meaning: words/phrases/documents are converted into vectors where geometric closeness = semantic similarity. When you finish capsules 02-03 and 06, this cheat sheet will help you review before exercise 07.

Next capsule: 02-what-is-a-vector.md

Next →2. What Is a Vector?