learn the ropes of modern AI.
Knowledge, generated and managed like code.
Learn is the human face of DaC — Documentation as Code, QuKaiZen's knowledge engine. Not a static glossary kept by hand, but a living, version-controlled corpus that is generated, verified, and managed the way software is: one source of truth, reviewed before it's promoted, served everywhere at once.
DaC organizes understanding into Worlds — subjects like AI, Astronomy, or Quantum. A World begins by being defined: the dictionary lays down its sourced, plain-language vocabulary — the substrate everything else stands on. From those fundamentals the World grows. Research, methods, and the latest understanding accrete on top, each piece referencing and building on the terms beneath it.
A World is a beginning, not an output — a subject opened by defining it, then developed until the knowledge is solid enough to bake into a model that owns it outright. What you read here is the first layer: the definitions everything else is built on.
Anyone can define GGUF. The dictionary shows you a real one. Each entry flips from term to a concrete example you can picture — the difference between reading about a thing and actually getting it.
This is a live slice of the dictionary — four real entries cycling front (the plain definition) to back (the worked example). The full collection covers fundamentals, architecture, training, fine-tuning, quantization, inference, formats, and QuKaiZen's own concepts.
Tap a dot to jump; tap the card to flip it yourself.
One plain-language sentence per term — no circular jargon, no prerequisites.
One concrete worked example per term, with real names and numbers.
Every term cross-links to related ones — follow the thread as far as you like.
The whole thing is machine-readable at the public /what API.
When did a glossary last give you a worked example? That's the difference between reading and knowing.
One growing, version-controlled source of truth — read by humans on the page and by agents through the API.
Every model-building term, defined plainly and grouped by category — fundamentals, training, fine-tuning, quantization, inference, formats, and more.
Each entry carries one concrete example with real names and numbers — the part that makes a concept actually stick.
Query any term programmatically: /what?term=gguf returns the entry; /what?all=1 returns the whole set.
The terms drifting through every explainer link straight here — /aidictionary#lora lands on the right card.
Super Skill, Wisdom per Watt, the Nucleus Seal — the suite's own ideas defined next to the field's standard terms.
New terms land as the field moves. One file; the page, the API, and the deep-links all move with it.
A term you can picture is a term you keep.
Most glossaries hand you a circular definition and move on. Learn hands you a definition and a worked example — and then gets out of the way. It's the front door precisely because it assumes nothing.
Free, open, and the same source whether you're a human reading the page or an agent calling the API.
per term — concrete, with real names and numbers. The part a glossary usually leaves out.
When did a glossary last show you, not just tell you?
Open the dictionary, pick a term you've always nodded along to, and finally see the example.