Lab

Visual grammar

Lab
A reference for visual decisions: what an image does, with which technique, backed by analysed campaigns.
Year
2026
Tools
  • React
  • TypeScript
  • D3
  • Claude Code

The idea

Images are usually argued about in terms of taste. What if every visual decision could be named: which effect it aims for, which lever carries it and where it typically fails?

What it’s for

A shared language for art direction, creative briefs and critiques. From insight to idea pattern, from pattern to the matching techniques, and on to examples showing how others solved it.

My own taxonomy of visual design as a local web app. It is organised in levels: five effects (differentiate, group, sequence, predict, infer), three domains, six levers and 51 techniques, each with a definition, when to use it, the typical mistake and related techniques.

The library holds 149 analysed campaign and poster images. Each image carries one to three techniques, marked as primary or supporting, with a rationale based on the subtraction test: what would be lost if the technique were missing? An agent proposes new images, and they are approved by hand.

  • Network: all images, techniques and effects on one map, filterable by lever, medium and industry
  • Taxonomy: techniques in a tree, with gaps where examples are still missing
  • Concept: 37 idea patterns in three groups, from insight to idea and on to the typical techniques
  • Compare: six to nine images side by side, saved as a set
  • Image sources: Ads of the World, D&AD Awards Archive, V&A Collections. The project only stores links, no image files
  • Structure based on «A Taxonomy of Visuals» by Modern Art Direction, idea patterns after Goldenberg, Mazursky & Solomon, Mario Pricken and Design Workout (Karpov). Library and rationales are my own