Brandon Sanderson was twenty-three years old and sitting on a pile of rejected manuscripts when he wrote Elantris. The magic system he built for it, called AonDor, let characters draw luminous symbols in the air to channel power. It was elegant, specific, the kind of system that makes you want to grab a pen and start sketching glyphs on napkins. But the actual engine of the novel had almost nothing to do with what AonDor could do.
It had to do with the fact that AonDor had stopped working.
Ten years before the story begins, the magic broke. The city of Elantris, once radiant and godlike, became a rotting prison. The people who were "chosen" by the magic still got chosen, still got thrown inside those walls, but the transformation that once made them luminous now made them something closer to undead. The whole plot, every political twist, every character's desperation, grew out of that one structural crack. A magic system that used to work and doesn't anymore. Sanderson didn't just design a system. He designed a system in collapse, and the collapse was the story.
He was young enough that I don't think he fully understood what he'd stumbled into. He later formalized the instinct into what fantasy writers now call Sanderson's Three Laws of Magic, which have become probably the most cited framework in modern genre writing. But the instinct came first. The framework came years later, reverse-engineered from something he'd felt his way through at twenty-three.
The thing most new writers get wrong about magic systems is they start with what magic can do. They build upward. More abilities, more variations, more spectacle. And the system gets bigger and bigger and flatter and flatter, because when magic can do anything, conflict has nowhere to live. Stakes evaporate. The protagonist waves a hand and the problem is solved.
The constraint is the story. Every interesting magic system in the history of fantasy literature is interesting because of what it costs and where it fails. It's the whole game.
Hard magic, soft magic, and the spectrum between them
Sanderson is the one who gave us the vocabulary. Hard magic has clear, defined rules that the reader understands. Soft magic stays mysterious, with limits the reader can feel but never fully map. The terms are useful. But the spectrum is much older than the terminology.
Tolkien wrote the most famous soft magic in the English language. Gandalf is a wizard. What can he do? He can fight a Balrog. He can break a bridge with his staff and a word. He can, apparently, die and come back as a stronger version of himself. But if you asked a reader to sit down and write out the rules governing Gandalf's magic, they'd stare at you. There are no rules. There's atmosphere. There's weight. You believe in Gandalf's power because Tolkien believed in it, and because Gandalf uses it sparingly enough that it always feels like something immense is being held back. The mystery is doing the work.
Ursula K. Le Guin's Earthsea sits in a nearby but distinct place. Naming has power. If you know the true name of a thing, wind or stone or another person, you can command it. That's a rule, sort of. But Le Guin never codifies it the way a hard magic writer would. You never get a chart of which names do what. Instead you get this pervasive sense that the world is held together by language, and that using a name recklessly could unravel something you don't understand. The cost isn't spelled out. It's implied. And it works because Le Guin was a writer who trusted implication more than exposition.
On the other end, Patrick Rothfuss built sympathy in The Kingkiller Chronicle as a system that practically runs on physics. You create a sympathetic link between two objects, and energy transfers between them, with loss. There's a concept called slippage, the energy that bleeds away in the transfer, and it obeys something close to conservation of energy. Kvothe has to do math in his head while someone is trying to kill him. The reader understands the system well enough to feel the tension in every scene where it's used, because you know exactly what Kvothe can and can't do. You know when he's outmatched.
Then there's Sanderson's own Allomancy from Mistborn. Specific metals grant specific abilities. Pewter enhances physical strength. Steel lets you push on nearby metals. Tin sharpens the senses. Every power has a defined scope. Readers can strategize alongside the characters, and the climactic scenes feel earned because you understand the toolkit. Hard magic at its most precise.
But most working magic systems, the ones that actually sustain novels, live somewhere in the middle of that spectrum. They have enough rules that the reader feels oriented and enough mystery that the world still feels larger than the plot. I'm not sure why this balance is so hard to talk about in the abstract. It's one of those things where you recognize it on the page but struggle to define it in a craft essay. The spectrum is real, though. And knowing where your system sits on it, even roughly, changes how you write every scene that involves magic.