Foundation-ness is precisely the degree to which learned representations remain coherent when the ground shifts; loss of coherence forces fallback to prior defaults.

{
  "doc": "/Users/antonborzov/Documents/Research.nosync/alexander/applied/alexander-and-levin.md",
  "claim": "Foundation-ness is precisely the degree to which learned representations remain coherent when the ground shifts; loss of coherence forces fallback to prior defaults.",
  "vector": "convergent-representations",
  "section": "Anton's \"Opus-style\" thesis statement",
  "extracted_at": 1778899506,
  "extracted_by": "haiku-4-5",
  "supporting_refs": [
    "Anton",
    "foundation models",
    "structural dharma"
  ]
}