The Text on the Table

On the table lies a text. You may know it well: the rules you have long played by, the instructions you work by, the theorem you remember from school. Beside it lies a model built from it. You can ask it a question, change its initial conditions, and check what follows from them. But first we need to see exactly what it has kept from the text — and learn to read it ourselves.

The rules of a game contain both an allowed move and a piece of advice to the player. A theorem’s exposition contains a conclusion and the conditions under which it is true. In a story, the hero may be sure of what the story itself does not confirm. In ordinary reading we tell such things apart without thinking. In a model these differences must be expressed explicitly: the answer depends on them.

This book teaches you to read such a model, change it, and finally write your own. You need no prior knowledge of the language, and no other books. First a statement and a question. Then the conclusion and its boundaries. Then sources, time, and interpretation. At the end — a model you can hand to another person.

One sentence

Let us start with a sentence from a general chemistry textbook, OpenStax Chemistry 2e, section 14.1. Water can give a proton to another particle, and it can accept one. The textbook calls particles able to do both amphiprotic.

That is all the chemistry we need for now. You need not remember what a proton is, only that it passes from one particle to another; there is no need to know any formulas. What matters is seeing three things in the sentence: what it is about, what it reports, and what follows from what is reported. Water. Gives a proton, accepts a proton. So it is amphiprotic.

The first four chapters will stay with this one sentence. We will write it down and ask questions about it. Then we will remove one piece of information and see what happens to the answer, add a contradictory one and see how the model keeps it, and read the answer’s explanation down to the line in the textbook. After that the subject will change, but the way of reasoning will stay: a triangle, a drawing, a mushroom, a parable, a standard, a unit of measurement, acid again. Each new step will build on the previous ones, and each time a real text will stand beside it.

Six questions for the text

Pick a small text you would like to look at closely. A familiar rule, some instructions, a reference definition, a short story. Choose something you can read in full and discuss in detail. Save it and open a sheet for notes beside it.

As we read, we will return to six questions. Answer them now in ordinary words, then check your answers against the chapters where each question gets its formal expression.

1. What is the rule here? Which phrases report information, which link conditions to a conclusion, and which convey someone else’s opinion or expectation? Chapters 2 and 5–8.
2. What is the text silent about? Which questions in it have no answer? How can we preserve that absence without turning it into a negation? Chapters 3 and 7.
3. Who decides? What follows from the recorded conditions, and what needs a separate decision by a human or an institution? Chapters 9 and 10.
4. What counts as a unit of the source? Does the text have articles, axioms, paragraphs? If you divide it yourself, why this way? Chapter 12.
5. Which version of the text is before us? What changed in other editions, and how can you tell which one your question needs? Chapters 13 and 14.
6. How can you hand the work over? What will another person need to repeat the check, trace the reasoning, and explain what they disagree with? Chapters 16–18.

Your answers can be provisional until you have met the language. They are landmarks. After each part, return to the notes: some answers will grow more precise, others will need revision.

How we will work

The material will be ready-made models from the Arxo corpus together with their sources and checks. A model in the book is always cited verbatim, with the file’s address; when an excerpt is shown, that is stated explicitly, and the full file is available through a link. The examples refer to a pinned historical version. The reader companion supplies the selected models, recompiles them with npm CLI 0.2.4, and records the adaptations needed for this edition. All printed commands and 383 selected scenarios have passed. The companion links to Syzdyq’s handbook without distributing its text; its source references are bibliographic rather than a local text snapshot. These checks do not establish equivalence to the retired runtime or replace independent reader feedback.

There are two ways to read. The reader’s route: predict the answer before it is printed, read the explanation after. The practical route: repeat the checks using the reader companion. Install the CLI below, then unpack the companion and open a terminal in its reader/ directory. The companion contains the models, dependencies, and checks; npm supplies the executable. Everything needed to understand the model is explained in the chapters.

Install the command-line tool

Install a current Node.js LTS release, including npm, from nodejs.org. The CLI requires Node.js 20 or newer. The published CLI supports macOS on Apple Silicon (ARM64) and Linux on x86-64. This edition’s execution checks were run on macOS ARM64. Install the version checked for this edition and confirm it is available:

node --version
npm --version
npm install -g @arxo/cli@0.2.4
law version --json

The last command should report "tool":"0.2.4". Keep that version for these examples; a later CLI may require different generated models or locks. If the shell cannot find law, add npm’s global bin directory to your PATH or use the alternative below. For a permissions error, follow npm’s user-owned installation guidance.

You can also run the same version without a global installation:

npx --yes --package @arxo/cli@0.2.4 law version --json

For a direct CLI command below, replace its initial law with npx --yes --package @arxo/cli@0.2.4 law if you choose this route. The question-preparation and receipt helpers additionally need Python 3 and law on PATH; the global installation above provides that command. No npm model package is needed: the companion supplies the book’s pinned inputs. Follow its README, and start each chapter in reader/. Keep commands that share shell variables in the same terminal session.

One warning about what this book does not do. It does not tell how the models were made, or where their authors went wrong. Before you lies the result, with checks for the properties we examine. Where a model covers only part of the text, that is stated; where the model’s author made a choice, the choice is recorded, and you can dispute it.

The grounds of knowledge

For us, the name Arxo is tied to the question of the grounds of knowledge. What holds up a statement we take to be true? What is given to us, what did we assume, and how did we reach the conclusion? In Grounds this question is explored through observation, evidence, and proof. Here we will walk one such path with our own hands: from a text we have read to a model whose reasoning can be checked.

Having a source does not spare us the work of interpretation. We choose word meanings, pick out conditions, and link phrases. These decisions are worth making visible, so another person can return to the same place and ask: does it really say here what we recorded? Then we can discuss a disagreement in concrete terms: find a missed qualification, propose another reading, pick a case in which the answers diverge.

Where we start

With water. The first chapter will open the textbook model’s vocabulary and read it line by line: the kind of objects, two relations, one statement. Then it will ask the first question and get the first answer with the grounds that support it. More complex ideas will come later, one at a time.

Let us open the model.