# shape (1-shape.org)

Shape is the outer form of something you can see or touch. People talk about the shape of a body, a cloud, a plan, or a story.

## What a model may hear
- tensor shape (machine learning frameworks like NumPy, PyTorch, TensorFlow): treat it as a tuple of dimensions (e.g., (3, 224, 224)) and may write code to reshape, broadcast, or validate array dimensions
- data shape / schema (databases, APIs, data engineering): assume the user wants to describe or enforce the structure of records, fields, or JSON objects
- geometric primitive (computer graphics, CAD, SVG): generate vector paths, meshes, or rendering commands for circles, polygons, or custom outlines
- SHAPE annotation (linguistics, NLP): refer to a morphological or syntactic annotation scheme, possibly triggering parsing or tagging behavior
- cloud instance shape (Oracle Cloud, some cloud providers): interpret as a VM configuration template with specific CPU, memory, and network resources

## Where people and models part ways
- Says: "Change the shape of this"
  Means: Make it look different
  May be taken as: Reshape a data array and break the data
  Say instead: "Describe the visual form you want, not using the word shape"
- Says: "What shape is this in?"
  Means: How good is the condition?
  May be taken as: Analyze tensor dimensions or schema structure
  Say instead: "Ask 'What condition is this in?' or 'How well is this working?'"
- Says: "Give it a better shape"
  Means: Improve the organization or flow
  May be taken as: Modify data structures or geometric meshes
  Say instead: "Say 'Reorganize this more clearly' or 'Improve the structure of the argument'"
- Says: "The shape of things to come"
  Means: A metaphor about future trends
  May be taken as: Attempt literal geometric or data analysis
  Say instead: "Keep the idiom but add 'as a metaphor' or rephrase as 'future trends'"

## Tips
- If you mean visual form, say 'outline,' 'silhouette,' or 'appearance' instead.
- If you mean data structure, say 'structure,' 'format,' or 'schema' to be clearer to both listeners.
- When discussing arrays or matrices, 'reshape' is the standard technical verb; use it only if you mean that operation.
- In cloud contexts, say 'instance type' or 'VM size' rather than 'shape' unless you specifically use Oracle Cloud.
- If you use 'shape' metaphorically, flag it: 'speaking metaphorically, the shape of this argument...'

## Often confused with
- form: More abstract; less tied to ML tensor dimensions
- reshape: A specific array operation; never metaphorical
- structure: Emphasizes organization over spatial or numerical dimensions
- figure: Can mean a number, a body, or a diagram; overlaps in geometry but not in data
- mold: A physical tool for shaping; no tensor meaning
- profile: A side view or summary; used in performance and user contexts