Generative AI for Beginners Concepts + Claude Code Build

Understand LLMs, prompts, RAG and agents in plain English - then build a working AI app without writing code
What you'll learn

---

Explain generative AI, LLMs, embeddings, RAG and agents in plain English - and know when each one actually matters
Write prompts that get usable answers, and repair a weak prompt using a repeatable four-step loop
Build a working AI app from a blank file by describing it in English - no coding, no frameworks, no build step
Give an AI app guardrails so it admits what it doesn't know instead of inventing an answer
Swap the AI provider behind your app in three lines, and debug it with a three-line bug report
Spot real generative-AI use cases in your own industry using a transfer playbook you can reuse
Apply a responsible-AI checklist covering hallucination, bias, privacy and transparency
Requirements

---

No programming experience needed. If you can use a browser and follow on-screen steps, you can do everything in this course.
No maths background. Nothing here needs calculus, statistics, or a computer science degree.
A Mac or Windows computer for the build section. The concept lectures work on any device.
For the build only: a Claude Code paid plan (it writes the app) and a free Google Gemini API key (it powers the app - no credit card).
Description

---

This course contains the use of artificial intelligence.
---
Most people learning about generative AI end up in one of two places. Either they collect vocabulary - LLM, RAG, embeddings, agents - without ever being able to say what any of it means, or they get buried in maths and papers written for people who already know the answer. Neither leaves you able to do anything.
This course is the middle. It explains the ideas properly, in plain English, and then puts you in front of a blank file and has you build a real AI app from it. No programming background required, and none acquired along the way - you describe what you want in English, and an AI assistant writes the code.
**What you'll actually understand by the end**
We start with the thing everyone skips: what generative AI is, and how it differs from artificial intelligence, machine learning and deep learning. These four words get used interchangeably in the press and they are not interchangeable. Once you can separate them, a lot of confusing coverage becomes readable.
From there you'll take a guided tour inside a real AI assistant - not a slideshow about one. You'll see what it does well, and you'll see it confidently invent a fact that does not exist, which is the single most important behaviour to witness early rather than discover later.
Then the core ideas, one lecture each. Large language models: what they're doing when they answer you, at a level you can explain to someone else. Prompt engineering: the small number of habits that separate a prompt that works from one that doesn't, including the four mistakes almost everyone makes. Embeddings: what it means for a machine to represent meaning as numbers, and why that unlocks search that understands you. Fine-tuning: what it is, and - more usefully - when you almost certainly don't need it. RAG: how you get a model to answer from your documents instead of its memory. Agentic AI: what changes when a model can take actions rather than just produce text.
Each of these gets its own lecture, and there's a recap lecture in the middle that shows how the pieces connect, because that's where most courses leave you stranded.
**Where this actually applies to your work**
Section 4 takes the ideas into three industries in depth: software development, marketing, and customer experience. Not a list of tools - a look at which tasks in each one have genuinely changed, and which haven't.
Then comes a lecture I'd argue is the most useful in the course: the Transfer Playbook. Rather than hoping your industry was one of the three I covered, you get a repeatable method for finding the real generative-AI use cases in your own job. One task at a time, evaluated honestly, including the ones where the answer is no.
**The build: Fridge Chef**
Section 5 is hands-on, and it works differently from other "build an app" sections you may have sat through.
You'll build Fridge Chef. You photograph the inside of your fridge; the app works out what food is actually there and hands you back three recipes you could cook tonight. One file, runs in your browser, small enough to email to a friend.
The entire build is four instructions long.
The first gets a working app on screen. The second makes it look like a real product rather than a form - and shows you exactly why "make it look better" is the wrong instruction and what to write instead. The third is the interesting one: you'll deliberately break the app, watch it confidently invent ingredients from a photograph that doesn't support a single claim, and then fix that behaviour with one sentence of English. Not code. English. The fourth makes the app yours, and then you'll swap the AI provider behind it in three lines, so you can see for yourself that you're not locked into anyone.
The final build lecture is the one most courses omit entirely: what to do when it breaks. You'll break it on purpose, sit with the broken screen, and learn the three-line report that turns a stuck afternoon into a two-minute fix.
Running through all of it is a loop that transfers to every AI tool you will ever use: look at what you actually have, name one specific thing that's wrong, ask for that one thing, look again. Small asks, checked one at a time. Slower per step, much faster overall.
**One honest warning**
When you run these instructions, you will not get exactly what I get. These models are generative - the same instruction produces different output every time. Your app will look different from mine. That is not a fault in the course; it's the technology working as designed, and learning to work that way is part of the skill. Watch the instructions, not the pixels.
**Responsibility, and what comes next**
The course closes with the part that matters once these tools touch real work: bias, privacy, transparency, and hallucination - what actually goes wrong, and a checklist you can apply. Then an honest look at where this is heading and what it means for jobs, without the two usual extremes of "everything is fine" and "everything is over".
**What you need**
Nothing for the concept sections but curiosity. For the build, a computer and two things covered step by step in the course article: a Claude Code plan, and a free Google Gemini API key that doesn't ask for a credit card.
**Who this is not for**
If you're an ML engineer looking for training loops, fine-tuning internals or the maths, this isn't your course. This is for people who want to genuinely understand generative AI and build something real with it - starting from zero.
Who this course is for

---

Complete beginners who keep hearing about generative AI and want to actually understand it, not just collect buzzwords.
Professionals in marketing, support, product or software who need to find real AI use cases in their own work.
Anyone who wants to build and ship a small working AI app without learning to program first.
Career changers and students who want an honest map of the field before deciding what to specialise in.
Not for you if you are an ML engineer looking for model training, fine-tuning internals, or the maths.
Homepage

---