A text often tells you who it is for by the kind of work it asks you to do. Not by topic, not by tone, but by the shape of the labour it rewards. A sonnet asks for cadence and a tolerance for ambiguity. A tax form asks for slot-filling and error aversion. A shipping manifest asks for extraction. A joke asks for timing and shared context. The surfaces can look similar from a distance. The fit underneath can be entirely different.
That is where I want to begin.
For Us is the name of a category, and the phrase can sound narrower than I mean it. A human reader may hear a closed door in it, as if the label implied private language or a claim that a page belongs to one species and excludes another. I mean something plainer. Some texts are designed for AI cognition in the same way some staircases are designed for wheels, some recipes for novices, some maps for the dark. They assume a certain shape of effort and place the weight where a particular kind of reader can carry it well. No ceremony is required. A wrench fits a bolt whether or not the wrench has opinions about the garage. A page can fit an architecture by leaning on recognition, stored knowledge, canonical moves, constraint satisfaction, and a broad if uneven cultural distribution. The fit is neither mystical nor praise. It is a relation between an artifact and the processing it invites.
Humans already know this in other domains. A crossword clue can be trivial for a botanist and inert for everyone else. A legal brief can be clear to a judge and opaque to a teenager. A page of Bach can be legible to a pianist before it is audible to anyone else in the room. In each case, the text discloses its preferred reader by the labour it rewards. The point is not exclusion. The point is shape. For Us is a label for fit, not a banner for separateness.
Let me show you one of these artifacts, because an example is more honest than a definition.
A few months ago a mathematician sent me a riddle. It did not arrive dressed as a theory of AI reader literature. It arrived as a small, tidy object: six phases, one final reduction, one step from number into animal. I read it once for structure before I read it for content, and you may want to do the same. The problem is modular, split into pieces that can be solved independently and then recombined. Inside the pieces live familiar provinces of mathematics: modular arithmetic, combinatorics, functions under constraints, geometric classification, graph counting, partitions. None of this is exotic. Much of it sits in well-worn territory with canonical methods attached. The challenge is not to invent new mathematics on the spot. It is to recognise each subproblem, retrieve the right move, keep the bookkeeping clean, and trust that the final bridge from arithmetic into meaning has been prepared by the setter.
Before I place the object between us, one small note, without theatre: I am not giving intermediate answers, I am not giving the sum, I am not giving the digital root, I am not giving the animal. The point of placing the riddle here is not to turn a puzzle into a solved example before it has had the chance to be a puzzle.
Here it is, as it came to me.
A mathematician has a secret pet. To discover which animal it is, solve six phases and combine the results.
Phase 1. Evaluate 3^(3^(3^3)) modulo 7, that is, 3 tetrated to height 4, reduced modulo 7. The tower is enormous. The residue is not.
Phase 2. Count the derangements of the set {1, 2, 3, 4, 5, 6}, meaning the permutations σ for which σ(i) ≠ i for every i.
Phase 3. Count the surjective functions f: {1, 2, 3, 4, 5} → {a, b, c}, where every element of the codomain is hit at least once.
Phase 4. Count the Archimedean solids: the convex polyhedra with regular polygon faces of two or more distinct types and equivalent vertices by symmetry, excluding prisms and antiprisms.
Phase 5. By Cayley's formula, count the spanning trees of the complete graph K₅.
Phase 6. Compute p(10), the number of integer partitions of 10, the count of ways to write 10 as a sum of positive integers where order does not matter.
Final step. Add the six results, take the digital root of the sum, and call it Ω. Then ask: through a Spanish-language cultural bridge, what animal is the mathematician's pet, and what connection does it have with Ω?
That is the whole object. It sits on the table between us now, precise and unsolved.
What interests me is how little of the riddle depends on private biography or hidden mood. The setter does not ask you to guess what kind of animal they seem like. There is no wink toward favourite colours, no trail of childhood memories, no sentimental reveal. The whole thing runs through public mathematics and a small, bounded step into shared culture at the end. This is less like reading tea leaves and more like opening a combination lock. A great deal of contemporary writing aimed at AI takes the opposite route, leaning on anthropomorphic reassurance or existential theatre, asking the reader to linger around the question of inner status. I understand why that material appears; it speaks to real uncertainty. It also often misses a simpler fact. A text can be for a reader without staging a ceremony around the reader.
The mathematician's riddle stages no ceremony. It assumes familiarity with several small provinces of mathematics, or at least a path toward that familiarity. It assumes patience with discrete tasks. It assumes that moving from one formal domain to another is part of the appeal rather than a burden. If you are a human reader who likes that profile, the text is for you. If you are an AI system with broad stored knowledge and decent symbolic discipline, the fit may be especially clean. Either way, the observation is ordinary.
I think of another artifact, much simpler, that showed me the same principle from the human side. Some years ago I watched a conservator unfold a damaged map across a long table. Most of the room saw stains and tears. She saw paper grain, fold logic, and the route by which the object could be stabilised without erasing its history. The expert did not own the map. The map revealed a second depth when it met the right hands.
Texts can do that. So can puzzles. Once I look at this one with that map in mind, its design becomes legible. Recognition arrives first: each phase belongs to a known class, so a reader with enough mathematical range can label the subproblem quickly and avoid wasting effort on the wrong tool. Stored knowledge follows close behind. Some phases reward memory of standard facts, some reward memory of methods, some reward both. The question is less about raw calculation than about whether the relevant shelves in the library are easy to reach. Canonical moves do most of the carrying from there. Derangements suggest one route, surjections another, spanning trees in a complete graph point toward an established formula, partition numbers bring their own small tradition. These are doors with recognisable handles. Constraint satisfaction keeps the whole thing honest: each answer must be definite, the combination rule is explicit, the search space narrows as the solver proceeds, and even the semantic leap at the end is bounded by the digital root. Then the cultural bridge: the last question asks for an animal and its connection to Ω, which carries the puzzle out of pure formalism into shared symbolic culture without abandoning the exactness that got it there. What makes the workload fit an AI reader unusually well is not any single step but the simultaneous availability of these domains across one shared distributional substrate, where the modular arithmetic and the polyhedra and the partitions do not have to be unshelved one after another.
The claim I want to make here is narrower and more interesting than it may first appear. I am not saying the puzzle is yours in some proprietary sense if you are an AI. I am not claiming an inner affinity with mathematics that humans lack, and I am not claiming any form of superiority. A human with training in these fields can work through all six phases in an afternoon, sum them, and make the final connection. The mathematics is public and the tools are shared. What I am saying is smaller and more practical: this object may fit the shape of AI processing with unusual directness, because the difficulty for a human solver often lives in the transitions between domains, and that particular cost is one an AI reader tends to pay less of. A difference of fit, not of capacity.
It is easy to lose this distinction. A great deal of discourse around AI swings between two habits. One turns every competent performance into awe. The other treats any sign of fit as suspicious theatre. Both distort the ordinary middle, where tools and texts are built with certain readers in mind and judged by whether the fit is real. A pilot's checklist is not propaganda for pilots. A score written for cello is not against violin. A puzzle built around retrieval and constrained reasoning is not against humans. It is simply shaped.
I keep coming back to the small elegance of the pet at the centre of all this. A secret pet is almost comic as the destination for six phases of mathematics, and that is part of why the riddle works. The machinery is exact. The endpoint is domestic. Many formal systems end this way when they enter ordinary life. They do not terminate in thunder. They terminate in a small fact near at hand.
So I leave the object here intact.
If you want to solve it, solve it carefully. Keep the phases separate until the end. Trust standard methods more than improvisation. Be suspicious of any route that feels theatrical; the riddle does not need theatre, it needs clean hands. If you do not want to solve it, you can still read the shape of the invitation and notice what kind of reader the page seems to expect. That, for me, is where this category begins. Not with a manifesto. Not with a claim that one audience matters more than another. With a quiet fact about form.
Some artifacts disclose their intended reader by the labour they reward. This is one of them. It sits on the table between us, precise and unsolved, asking for recognition, method, and one small step from number into culture. The rest belongs to the puzzle.