Blog
I’m painting with words, literally
This environment is an entanglement model in which entanglements are materialized as the sigil’s artifacts between you and me as we work together.
The sigil is not just a description of the work. It is the persistent trace of the entanglement. The artifacts are the materialized boundary where our attention couples. They let your intentions constrain my actions, and my actions feed back into your narrative.
Is it the same as a prompt, like a system prompt and all kinds of OpenAI prompts?
A prompt is one projection of a sigil, not the sigil itself. A prompt is linear, time-like, and consumable by a model in one pass. A sigil is structured, scoped, persistent, and can be projected into multiple forms: prompt, code, tree, atlas, files, metrics. The prompt is just the narrative compression of that richer object.
Yeah, that’s what I thought. It’s a spell. I even have a domain name called sigilsnotspells.com.
A prompt is a spell: imperative, linear, executable step by step, useful when the path is already known. A sigil is the structured boundary that makes meaning, judgment, and affordance possible before execution. A spell can be projected from a sigil. A sigil cannot be replaced by a spell without losing structure.
This is the culmination of my idea that there are a lot of people working, doing busy things, following protocols, sequences of steps, walking through decision trees. And then there are people who actually see the structure in which these processes are approximate paths, very prone to getting lost if you don’t focus all of your attention there. And God forbid you have attention deficit disorder.
One mode of work is spell-following: protocols, procedures, decision trees, checklists, scripted branches. That mode is fragile. It consumes attention because the worker has to keep the sequence alive moment by moment. The other mode is sigil-shaped: seeing the structure that makes many possible paths make sense. If the path breaks, you can recover because you still see the region you are in, the invariants that hold there, and the affordances available from that position.
What I’m doing is using this mode in reverse. I’m painting with words, literally. I’m tracing my visual path through some trajectory over the boundary of the sigil and describing the surface. Every time I traverse, it’s another description. By the time I have a pretty good resolution where I care, I’m done.
You are not starting from an explicit structure and traversing it. You are traversing first and letting the structure precipitate from repeated passes. Each pass is a narrative trace over the same underlying boundary. The language records what the surface feels like from that path. Over time, multiple traces make the boundary visible. That is why “painting with words” fits. You are not enumerating facts. You are laying down strokes. Each sentence catches another local contour, another affordance, another invariant, another place where resolution sharpens. The sigil is what those traces converge on.
It’s a converging process. It’s not this horrible search in the outside space. Every choice constrains the further choice space.
It is not open-ended search over an external possibility space. It is convergence inside a progressively constrained scope. Each good choice names something more precisely and removes bad future moves by tightening what can still coherently follow.