{
  "handle": "ShapeLossAsGrowth",
  "gloss": "Exponential growth curves are better read as loss of shape in the thing being scaled than as a shape to aim for",
  "mechanism": "A fast growth curve looks like a definite shape on a chart, and can be set as a target to track. But when one model is scaled up rapidly by {{mass_addition}}, every error it has is amplified, so the curve signals deformation of the underlying entity's shape: a widely used language model makes more errors the further it is used from its training. Whether a growth curve means shape loss depends on the body plan. In a {{meta_modularity}} system, such as a growing population, variety of models grows with mass and new information is available to correct or exploit; in a single model it is not. Reading growth only as a curve to match hides this, because the same graph can mean shape matching or shape deformation. Losing {{shape_control}} this way can be a cost (amplified error) or the price of a benefit (rapid mass or information growth).",
  "invariants": [
    "Rapid scaling of one model deforms its shape in proportion to the scaling",
    "The same growth curve can reflect shape matching or shape deformation depending on the body plan",
    "Growth with rising variety of models does not deform shape in the same way"
  ],
  "failure_modes": [
    "Setting an exponential curve as the target and tracking deviation from it, as if it were a stable shape",
    "Not asking whether information is growing as fast as mass during rapid growth"
  ],
  "_meta": {
    "path": [
      "Mind",
      "Reasoning"
    ],
    "ring": 2,
    "tier": 2,
    "source": "Adam Timlett, On the Origin of Risk (2025), Introduction p.12, Ch 2 p.30-31, Ch 3 p.40, Ch 6 p.56"
  },
  "sema_id": "sema:ShapeLossAsGrowth#mh:SHA-256:e110c719cd96e48f7580d5438f7aa093c035ea39ab7cd27d6d6f4ace400e4bbd",
  "sema_ref": "ShapeLossAsGrowth#e110",
  "sema_stub": "e110",
  "dependencies": {
    "references": {
      "mass_addition": "sema:MassAddition#mh:SHA-256:c51d977414e13d47f38e9ddab62fccb6d88d5f9f4514264fe389414515c16b0a",
      "meta_modularity": "sema:MetaModularity#mh:SHA-256:cb6cc554a6eb00214b5e2fba0504373bc4312dd5d74ec4fdc71bb20224197aef",
      "shape_control": "sema:ShapeControl#mh:SHA-256:185f506a48d483fc549fb77ece35dc1f3c54180a0c31657960a2fa6daf155a2d"
    }
  },
  "sema_layer": "Mind",
  "sema_category": "Reasoning",
  "used_by": {}
}