Feature Library
Free reading

Playbooks. How the parts are made, and why.

Each one is what I learned on real projects, with the reason for every rule and a checklist to run at the start of your own work.

Page shapesThe overall shape of a page before its parts: what the first screen does, the kind of header, one column, sections, cards or a sidebar, and where the reading ends, for nine kinds of page from an app workspace to an event, and a method to compose a shape when none fits.Placement and layoutWhere menus, toolbars, panels, charts, forms and buttons go, and why: one home per job, functions grouped by what they act on, editors beside their work and never over it, things that grow kept to free room, settings panels in rows, live figures that say what they are, and a layout that holds from a phone with its keyboard open to a wide desktop.Layout studioSketch a screen on desktop, tablet and phone grids from a template: move, resize, swap and rename its blocks while the placement rules are checked live, then copy a prompt, CSS and JSON for an agent to build it from.Performance budgetsA frame rate turned into milliseconds, shared out before the work starts: budgets for the worst frames, one allowance for work that can wait, counts per device tier, nothing drawn while nothing changes, and measurements you can trust.Motion and feelWhen to move things and how: motion that explains a change, a few named durations and curves, changes animated but arrivals not, springs that feel the same at any frame rate, cheap and interruptible motion, and reduced motion done properly.Accessibility baselineWhat every screen should meet before it ships: keyboard first, contrast tested on the theme, nothing said by colour alone, names and states, changes announced once, charts with words, and zoom that holds, with the checks to run.Layouts that never breakComponents that hold whatever goes in them: long names, translations, huge numbers, missing pictures, right to left, text at 200% and little room, with every state designed, sizes that bend within limits, and the tests that find what spills.Ingest that never double-countsReading files from a folder into a sheet or a database without ever writing one twice or losing one: know each file by its content, never write when the duplicate check can’t run, keep your own record of what was written, wait for files to settle, one writer, a log nobody can rewrite, and two PCs sharing one synced folder.Sculpted, not assembled: 3D assets from a Blender scriptBuilding a model in code so it reads as one sculpted form: overlapping volumes unioned, voxel-remeshed, smoothed and decimated per tier, shaded from a smoother copy, crisp lines cut last, review renders that resume, a ray sweep for parts showing through, and names a runtime can trust.Office and desktop companionsTaking web work where office people already are: a live web page inside a PowerPoint content block (how the block stores it, what its browser allows, no network, storage sealed per block, the white first frame, focus handed back to the clicker), a local web app shipped as a Windows tool (PyInstaller and a per-user Inno Setup installer, an Edge app window, one instance, a tray icon, start with Windows), and rows appended to an Excel workbook without ever breaking it.Many agents, one codebaseRunning several AI agents on one repository at the same time without them colliding: how to shape the code so they can’t, what goes in their brief, and how to check what comes back.Branching and forkingWhen parallel work has to touch the same files: a branch and a worktree per worker, merging them one at a time, and when an AI helper should be a fork, a fresh agent or a session of its own.Several people, each with their own agentsWhen a team shares a codebase and everyone brings their own AI helpers: the agreements that have to live in the repository, who claims what, who may approve what, and how work survives the end of a session.