About Spark Lamp

Why this site exists

AI is changing how we work and live, yet most explanations either stay vague ("it's like a brain") or jump straight into equations. Spark Lamp sits in between: it explains how modern AI really works, from the chips it runs on to the language models you chat with, in plain words and without assuming any math background.

Who it is for

Complete beginners who are curious about AI and want to reach a solid, advanced understanding: students, developers moving into AI, product people, and anyone who wants to know what happens behind the chat window.

How the courses work

• Read. Each module explains one idea in a few short paragraphs, with every technical term explained when you hover over it.
• Try it. Every module has its own interactive simulation with a small task, so you see the idea at work instead of taking it on trust.
• Check. A short quiz confirms you got it. For the curious, a "Go deeper" section adds the technical details.

What's inside

The courses form one path, each building on the ones before it:

• How Do Machines Learn? (Track 1): What AI is, how it differs from ordinary software, what a neuron does, and how a network learns from its mistakes.
• How Does ChatGPT Actually Work? (Track 2): How a chatbot works inside: tokens, meaning, attention and the transformer, how models are trained and scaled, and why they hallucinate.
• Why Does AI Run on GPUs? (Track 3): Why AI needs so much simple arithmetic, how GPUs and tensor cores do it in parallel, why moving data is the real bottleneck, and what it takes to run and train a model.
• How Does AI Serve Millions of People? (Track 4): What happens after you press send: prefill and decode, batching, the KV cache, splitting models over GPUs, serving engines, and what each answer costs.
• How Does an Application Ask an AI Model? (Track 5): How apps talk to models through APIs, how to choose a model, how to write prompts and system prompts, and how to get structured, streamed and affordable answers.
• How Do You Give AI Its Own Knowledge? (Track 6): Why models don’t know your facts, how meaning search and vector databases find them, how RAG feeds them to the model, and when fine-tuning is the better tool.
• From Chatbots to AI Agents (Track 7): How a model uses tools, works in a loop of thinking and acting, remembers, connects through MCP, works in teams, and stays under your control.
• AI Beyond Text: Seeing, Hearing and Creating (Track 8, coming soon): How models understand and create images, speech and video, not only text.
• Can You Really Trust AI? (Track 9, coming soon): How to judge an answer, where bias and leaks come from, how models get tricked, and whether we can see inside them.

Next to the courses, the Lab offers hands-on workshops you can do any time, such as Neural Network Foundations, where you build and train small neural networks yourself.

Who makes it

Spark Lamp is created and maintained by SparkLamp. Questions, corrections and ideas are always welcome on the contact page.