Hands-on AI training, built on visualizers you can break.
AI stopped being a demo. It's the tab open at work, and nobody had to explain why.
Most people learn a handful of prompts and stop there. The model stays a magic box, so when it fails, you learn nothing from it.
Nobody needs a math degree for this. The ideas just need to arrive in the right order, with something to click at every step.
The people who understand the engine will always get more out of it than the people who only know the pedals.
Every path is written start to finish by one person, so the difficulty actually climbs instead of resetting every module. Start at the fundamentals, or skip straight to the topic you came here for.
BeginnerWhat AI is, how it reasons, why it works.
4 modules · 21 lessons →
IntermediateDeep neural networks, CNNs, RNNs, generative models.
4 modules · 20 lessons →
AdvancedFrom text to language models: tokenization, attention, GPT.
5 modules · 25 lessons →
IntermediatePrompts, APIs, RAG, agents, MCP: build real AI systems.
7 modules · 35 lessons →
AdvancedAttack and defend AI systems: injection, jailbreaks, agentic risks.
6 modules · 29 lessons →
BeginnerEthics, literacy, meta-skills, the human edge.
3 modules · 15 lessons →Every lesson you finish pays you Compute, the same currency that runs the AI tools on Gradivex. The more you learn, the more you can build.
Work through a lesson and pass its end check. Real understanding, verified before you move on.
Each lesson and module pays out Compute and XP. A perfect score pays more.
Put Compute into real AI labs and challenges. That practice teaches you the next thing.
Ask questions, break things, tell us what is confusing. Half of what is on this page came from one conversation with an early user.
100% free while we're in early access, no credit card