Look closely at what is real.
Magic begins with a wish, but it only works after seeing the people, the work, the unspoken judgement, and what must already be protected.
The Sage of Beginnings · 始まりの賢者
KoichiNishizuka
Making magic real.
I turn unrealised possibilities into systems people can choose, verify, and reverse. My work is to build a responsible first path into the unknown.
First principles
The age of magic begins with responsibility.
As intelligence and power grow, we must first decide who chooses, what the power may touch, and how the result can be known.
Leave the final choice with a person.
AI can unfold many possible worlds. The responsibility for choosing which one becomes real must remain with people.
Keep power within a visible boundary.
Grant one permitted effect, not a key to everything. The stronger the power, the clearer its limit should be.
Leave proof and a way back.
Independently verify what changed, connect the choice to its result, and let a person choose to undo it when needed.
Work / Working evidence
Turn a principle into a tool that can touch the world.
Magic is the work in progress at the centre. Spell Runtime is public evidence that trust, boundaries, verification, receipts, and reversal can exist in working code.
01
Magic
A system for turning unrealised possibilities into exact plans, final human choice, bounded effects, independent verification, evidence, and a way back—without making AI the authority.
In development · private product workspaceCreator and system designerPublic summary only02Spell Runtime
A public runtime that inspects portable SpellBundles, checks signature, trust, policy, and risk state before execution, and records a receipt as evidence.
v1.2.5 public releaseRuntime design and implementationGitHub source ↗Now
How much of the magic is real today?
I can build an initial working version. The next proof is that someone encountering it for the first time can make their own choices, use it again the following week, repair it when needed, and pass it on.
Start something new without exaggerating it.
I study friction in real work and the judgements that have yet to be expressed. I turn them into context, procedures, and tools that people and AI can use while keeping final authority with people.
I would rather publish working evidence and state its limits plainly than claim to have finished a universal system.
Bring one unanswered question.
Bring one real-world question, read the thought behind the work, and inspect what already runs. This is the first door.