Where to Go Next

You've seen some of the ways Field can work with code, images, sound and data. Now pick something you want to make. This page maps the documentation and suggests a place to start in each section. If you haven't taken the Tour yet, that's a good place to begin.

Looking for a particular kind of work? Try drawing from data, graphics on the GPU, or AI and agents. To find a call by name, go straight to the reference. For a project that crosses several of these sections, start with the chains below.

Chains across subsystems

Each chain page follows one piece of work through several parts of Field. They are the quickest way to see how the sections below fit together before you pick one to learn.

• A Polytope, Cut Open and Traced is the one to read first: a solid is built, cut in half and rendered.
• A Cellular Automaton You Can Hear goes the other way, from a Wolfram rule to the shortest pattern notation that plays it, rendered to a .wav.

The rest cross physics, patterns, RectSL, photographs, λ-terms and lighting; the overview lists the full set.

Working in a document

The document itself: moving around it, finding things in it, and getting back what it used to say.

• Moving Around the Canvas is the one to read first: panning and zooming, Cmd Cmd-and-hold to see every box an arrow key can reach, and the function keys that jump to saved views.
• File Management is short and saves a lot of time: "data.csv".file resolves inside the document folder first and the workspace project second, so a path keeps working after the document is moved or handed to somebody else.
• Your History with Field says which of Field's kinds of history holds what, and which tool opens each. Document History is whole-document recovery from saved versions; Code History shows what this line said a few saves ago, without leaving the editor.
• Snapping is a toggle rather than a mode, for a moment of layout cleanup. The drawing layer over the canvas is on the Tour, in Drawing and Notation.
• The Semantic Index adds search by meaning to the workspace search on /welcome, which otherwise matches words in document metadata and box text.
• Profiling is the page to open when a piece stops keeping up. Its panel records how long each box took, so you can see which one is slowing things down.

Seeing values

The value a box ends with is shown beneath it. These pages cover where that output appears and the ways to look at it.

• Box Output explains where printed values and display panels appear. Visualizers & Controls goes on to .h, sliders and the value visualizer.
• Interactive Guides shows how boxes can walk a reader through a document with slide notes, selections, callbacks, templated values, per-slide chat and callout lines.
• DataFrames in a Box works with typed tables; Charts as FLines turns data into plots that can sit alongside your other geometry. Both take over where the Tour's Looking at Data stops.

Drawing from data

• Linear Algebra comes first because everything after it uses it: the vector, matrix and quaternion vocabulary underneath movement, cameras, meshes and spatial data.
• Functions Examples turns Functions stacks into graphs and parametric FLines, and Functions Domains covers the domain algebra that governs how an Fn is sampled.
• Values Over Time draws interpolants, KeyframeScope parameter and code morphs, rolling correlations, and Kalman and particle-filter estimation, all as FLines or live plots.
• Signal Examples keeps raw signals next to filtered and event-detected ones, then turns to what they are made of in frequency: spectra, spectrograms, mel bands and resonators.
• Colors Examples compares colour helpers on actual geometry.

The geometry, sound and audio-AI groups in that section are larger than this list suggests; open the section itself for the full set.

Networks and dynamics

• Sorting Networks Examples draws fixed comparator plans, disorder decay, and annealing histories.
• Interaction Nets Examples shows reducer rules as before-and-after geometry.
• Boolean Networks Examples connects wiring diagrams, state spaces, attractors and rhythm grids, and Boolean Network Music Examples maps those steps into NoteProcess and live Dirt triggers.
• Neighbors & Graphs Examples draws nearest-neighbour queries, geometric graph families, 3D graph viewers and pathing overlays, all on one point cloud.

Small languages

• Querying Object Graphs and Datalog in a Box are a pair, read in that order: the first finds every value of a kind inside a live object, such as every filter cutoff in a synth, and the second answers questions that follow connections as far as they go, such as which boxes run downstream of this one.
• FieldML Cheatsheet and FieldML Examples cover Field's small ML-flavoured expression language, from its prelude of numbers, lists and vectors to boxes that generate and animate drawings.
• StackSL Cheatsheet and StackSL Examples document the concatenative stack language and its worked drawings, and RectSL Cheatsheet and RectSL Examples do the same for the rectangle-stack drawing language.
• ForthField Examples covers Field's own Forth, its output helpers, and its bridge into JVM objects. Nano Examples covers the tiny assembly-like machine and its RAM and register visualizers.
• Python in a Box and Clojure in a Box are the odd ones out: neither is small, and neither is Field's. Both run inside Field's own process, so a Python cell can import java.lang, reach its own Box, and have fig.show() draw into that box instead of a browser.

The section ends with pages about making your own. Rewriting Your Own Syntax is the lightest: a rewrite { … } block in a box matches a shape in your source and rewrites it before the compiler sees it. Bridging Another Language brings a language in from outside. Language Extensions is how any of them earns highlighting, diagnostics and a context lens inside the editor.

Graphics, GPU and headsets

The Tour's Graphics on the GPU stop gets a shader drawing in a window. These pages go further.

• Field 4 Graphics goes underneath that, step by step: triangles and buffers, cameras, compute shaders, textures, glTF models and render-to-texture.
• Fabricate Renderer covers the StandardRenderer bundle: scene graph, render planning, frame graph, post stack, glTF import, material response, shadows and ambient occlusion.
• Aperture Playground is the worked multi-pass example: light bands, reductions, aperture filters and a composite, each in its own box, with every shader stage in the document. It shows defocus spreading a line rather than just dimming it.
• Field Probes instruments live shaders, with the CPU read-back fieldProbe and the GPU-resident fieldStats both demonstrated on one compute-shader particle system.
• Apple Vision Pro is Field's app for the headset, still being built: it puts the editor and canvas in the room beside 3D geometry and the renderer's output. Geometry in the Headset is written for code running inside that app; only its geometry half can be previewed on your Mac.
• Ray Tracing Examples walks through VK_KHR_ray_tracing_pipeline programs, from a raygen shader writing into a storage image to procedural spheres reflecting a traced floor. They need a GPU and driver with hardware ray tracing; the path tracer has a compute backend for machines without it.
• Kotlin in the Head runs a box's Kotlin in a browser tab or on the headset instead of inside Field, for 3D scenes, sound and shaders.

AI and agents

• AI Overview is the one to read first. It is organised by what you are asking for: a conversation about a box, an edit at one spot, AI candidates mixed into ctrl-space, or a function body filled in from its signature.
• Agent Transcript Gallery collects agent-built documents as reproducible starting points, each materialized from a recorded transcript and openable in your own workspace.
• AI Tool-Use History is the structured trace of what a send actually did, for when a tool call goes wrong, and AI Documentation Polish covers the generated documentation that fills in for a completion entry whose source carries no useful KDoc.
• Notes on the Documentation turns notes you leave on these pages into a queue that agents work through.

Two pages you might expect here are on the Tour instead, because both are about reaching a running document rather than about Field's own AI: Automating Field, on the macros, the transcript format and the backdoor script that let a script or an agent drive Field the way a person does, and Field from your phone, which turns /controllers into a phone-sized remote for live controls, outputs, AI4 chat, Open Anything selection and Monaco editing.

Reference, and internals

Neither section is a worked example. The reference is for looking something up; the internals are for understanding how Field works underneath.

What's Already in the Box lists every public declaration in Field's standard library, grouped the way the editor's autocomplete groups it, with Used in links back to the pages that use it. It is the fastest way to find out whether something already exists. Beside it is the Generated Vector API, emitted from the same model that emits Field's vector types.

Internals is the other extreme: hand-written prose about mechanism, for when you are extending Field or working out why it did something, rather than writing a box. What Field Exposes is the one to start from: the HTTP and WebSocket surface your own editor already talks to. What a Document Is on Disk is the one to read before editing a document by hand, because it says which file in that folder wins when Field loads.