This is an offline app featuring 51 hands-on interactive experiments designed to help you deeply understand how AI actually works under the hood—covering everything from individual neurons to full-fledged agents…
Key Highlights
- Features 51 experiments across 6 categories, covering everything from fundamental concepts (what is a neuron, gradient descent) to attention mechanisms, Retrieval-Augmented Generation (RAG), agents, fine-tuning, quantization, model deployment & serving, and much more.
- Each experiment comes with a concise tutorial supporting two reading modes: “Plain” and “Standard”.
- Some experiments are genuinely live and interactive, not just pre-rendered animations:
- The attention mechanism experiment runs a trained, lightweight Transformer model (1.37M parameters) right inside the app.
- The training experiment trains a tiny character-level model live in real time as you adjust the sliders.
- Completely offline: No account sign-up required, zero telemetry data collected.
How It Was Built
I selected the topics, conducted the background research, and designed the overall structure and taxonomy. The tutorial texts were co-written with Claude and GPT Astra, with Grok used for a second pass of source-checking and verification. There may inevitably be mistakes—if you spot any issues, feedback and pull requests are very welcome!
If you teach in a related field or work in a specific domain of AI, I’d love to collaborate on expanding it: whether that’s adding a new interactive experiment, designing better visualizations, or refining existing experiments to make the causal logic clearer. I’d also love to hear which experiments felt confusing or what topics you think are currently missing.
GitHub: https://github.com/Fazmin/AILearningGuide
Top Discussions
dan-lash (⬆13): Thanks for making this! I’ve always been interested in fine-tuning but never knew where to start—even if you already know a bit, rebuilding your understanding from the fundamentals up is definitely the right way to go. Zipidyzip (OP) (⬆2): Thanks! That was exactly where I was at too. I understood certain pieces well, but knew very little about others. Not sure if you noticed, but there’s a place in the app where you can see how all the pieces connect together (the Connection tab), allowing you to visually explore the relationships between different modules. The_VibePreneur (⬆12): Actually running a tiny model inside the app rather than just showing a canned simulation animation is such a great touch. Bookmarked! This is perfect for onboarding newcomers—especially those who want to grasp the actual inner workings instead of just chasing buzzwords. Zipidyzip (OP) (⬆2): Thanks a lot! Great to see it being useful in different scenarios :-) Columnexco (⬆5): Absolutely love this! My only suggestion would be to make the text a bit more accessible and plain-spoken. Zipidyzip (OP) (⬆2): Totally fair point, thanks for the feedback! I tried to tackle that with the “Plain” and “Standard” modes, but the “Plain” mode might still feel a bit dense for some folks, especially in the more technical sections. Were there any specific parts that felt too arcane to you? Maybe down the road, I could offer an even simpler, story-driven version with concrete examples for certain modules. LevelBruce (⬆8): The backprop diagram with multiplication nodes hits the sweet spot for guided walkthroughs—super helpful for anyone who’s only ever seen pure mathematical formula derivations. Zipidyzip (OP) (⬆6): Thanks! Haha, I figured if I was going to put people through the torture of the chain rule, the least I could do was give them a node to hold onto :-)