[
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq3answer",
    "@type": [
      "http://schema.org/Answer"
    ],
    "http://schema.org/text": [
      {
        "@value": "A top white-shoe M&A practice runs ~1,000 deals per year with practice-area-specific deal shapes (NDA → term sheet → diligence → purchase agreement) and confidentiality constraints. Transformation requires an operator with ambiguous intermediate goals, incomplete feedback, and a multi-year horizon — no single eval can capture it."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#concept_tokenEconomy",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "As articulated by Matt MacInnis at Rippling: a token answering a generic question is worth almost nothing; a token reasoning over proprietary company data is worth much more."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Token Value Gradient"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#evidence_openEvidence",
    "@type": [
      "http://schema.org/Observation"
    ],
    "http://schema.org/about": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#org_openEvidence"
      }
    ],
    "http://schema.org/description": [
      {
        "@value": "A majority of American doctors now open OpenEvidence every day. No amount of compute buys physician habit or decision-flow integration; trust is built slowly through relationships."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "OpenEvidence Physician Adoption"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_rippling",
    "@type": [
      "http://schema.org/Organization"
    ],
    "http://schema.org/name": [
      {
        "@value": "Rippling"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://www.rippling.com/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#concept_privateTruth",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "Correctness that exists only inside a firm's data, workflows, and accumulated judgments. Neither trainable from public data nor verifiable by external benchmarks."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Private Ground Truth"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#claim_intelligenceNotBottleneck",
    "@type": [
      "http://schema.org/Statement"
    ],
    "http://schema.org/description": [
      {
        "@value": "A model smarter than any person still needs to be let in the door. Capability does not confer license, liability, or ownership of private data. Trust is built slowly on relationships, not gradient descent."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Intelligence is not the bottleneck — permission and accountability are"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#gloss_groundTruth",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "The authoritative correct answer to a question. 'Private ground truth' exists only inside a firm's systems and cannot be accessed or verified externally."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Ground Truth"
      }
    ],
    "http://www.w3.org/2002/07/owl#sameAs": [
      {
        "@id": "http://dbpedia.org/resource/Ground_truth"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#gabePereyra",
    "@type": [
      "http://schema.org/Person"
    ],
    "http://schema.org/description": [
      {
        "@value": "Observed that real automation requires the product, model, workflow, and firm to move together — and three of those four move at the speed of an organization."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Gabe Pereyra"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#gloss_verifier",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "A check that costs nothing to run — a compiler, a test suite — enabling rapid grinding against a benchmark. Its absence is what makes a task illegible and resistant to training."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Free Verifier"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_baseten",
    "@type": [
      "http://schema.org/SoftwareApplication"
    ],
    "http://schema.org/description": [
      {
        "@value": "Inference serving provider; best AI-native companies concentrate serving on one or two providers for reliability and guaranteed compute access — not just cheapest token price."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Baseten"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://www.baseten.co/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#step4_priceOutcomes",
    "@type": [
      "http://schema.org/HowToStep"
    ],
    "http://schema.org/name": [
      {
        "@value": "Price outcomes, not inputs"
      }
    ],
    "http://schema.org/position": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#integer",
        "@value": 4
      }
    ],
    "http://schema.org/text": [
      {
        "@value": "Charge for resolution, not tokens. Outcome pricing is only credible if you own the definition of success and are trusted inside the system. It signals depth of integration."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#step5_avoidCapitalWar",
    "@type": [
      "http://schema.org/HowToStep"
    ],
    "http://schema.org/name": [
      {
        "@value": "Avoid the general-model capital war"
      }
    ],
    "http://schema.org/position": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#integer",
        "@value": 5
      }
    ],
    "http://schema.org/text": [
      {
        "@value": "Train specialized models on your private data and evals for narrow tasks. Companies that try to out-train frontier labs on general swaths usually end in a sale to someone compute-rich."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#mattMacInnis",
    "@type": [
      "http://schema.org/Person"
    ],
    "http://schema.org/affiliation": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#org_rippling"
      }
    ],
    "http://schema.org/description": [
      {
        "@value": "Rippling executive who articulated the token value gradient: generic-question tokens are near-worthless; proprietary-data-reasoning tokens carry real value."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Matt MacInnis"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#concept_absorptionFrontier",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "The expanding boundary at which model labs pull application-layer scaffolding (retrieval, routing, tool use, reasoning) into the base model weights, eroding the margin of app-layer companies."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Absorption Frontier"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#noamBrown",
    "@type": [
      "http://schema.org/Person"
    ],
    "http://schema.org/affiliation": [
      {
        "@id": "https://openai.com/#this"
      }
    ],
    "http://schema.org/description": [
      {
        "@value": "OpenAI researcher, pioneer of reasoning models. Observed that a one-year-horizon agent may require a full year of real-world operation to evaluate."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Noam Brown"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#evidence_noamBrownQuote",
    "@type": [
      "http://schema.org/Quotation"
    ],
    "http://schema.org/author": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#noamBrown"
      }
    ],
    "http://schema.org/description": [
      {
        "@value": "OpenAI's Noam Brown wrote that the only sure way to evaluate an agent over a one-year horizon may be to run it for a year — the clock cannot be skipped."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Noam Brown on Evaluation Horizon"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#evidence_devinBenchmark",
    "@type": [
      "http://schema.org/Dataset"
    ],
    "http://schema.org/description": [
      {
        "@value": "Devin shipped in 2024 solving 13% of SWE-Bench tasks. By mid-2026 the best agents hit the high 80s. The benchmark was conquered but the 'wrong lesson' was drawn — engineering resists full measurement."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "SWE-Bench Devin Trajectory"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_cognition",
    "@type": [
      "http://schema.org/Organization"
    ],
    "http://schema.org/description": [
      {
        "@value": "AI software engineering company offering a 'performance guarantee' on Devin's coding output — only possible because they're trusted inside the customer's system."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Cognition (Devin)"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://cognition.ai/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#article",
    "@type": [
      "http://schema.org/Article"
    ],
    "http://schema.org/about": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#thesis_untrainableValue"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#concept_absorptionFrontier"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#concept_illegibleCorrectness"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#concept_privateTruth"
      }
    ],
    "http://schema.org/author": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#sarahGuo"
      }
    ],
    "http://schema.org/dateModified": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#date",
        "@value": "2026-06-10"
      }
    ],
    "http://schema.org/datePublished": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#date",
        "@value": "2026-06-10"
      }
    ],
    "http://schema.org/description": [
      {
        "@value": "A thesis-driven investment essay arguing that durable AI value lies not in model capability but in private ground truth, domain trust, and illegible correctness that no benchmark can reach."
      }
    ],
    "http://schema.org/headline": [
      {
        "@value": "The Untrainable"
      }
    ],
    "http://schema.org/keywords": [
      {
        "@value": "AI investment, untrainable, private ground truth, absorption frontier, illegible correctness, benchmarks, AI moat"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "The Untrainable"
      }
    ],
    "http://schema.org/publisher": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#saranormousSubstack"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#gloss_distillation",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "Technique to produce a smaller, cheaper model from a larger one. Enables focused apps to run a workflow on a fraction of the token spend of a general agent."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Model Distillation"
      }
    ],
    "http://www.w3.org/2002/07/owl#sameAs": [
      {
        "@id": "http://dbpedia.org/resource/Knowledge_distillation"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#step1_getInside",
    "@type": [
      "http://schema.org/HowToStep"
    ],
    "http://schema.org/name": [
      {
        "@value": "Get inside one"
      }
    ],
    "http://schema.org/position": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#integer",
        "@value": 1
      }
    ],
    "http://schema.org/text": [
      {
        "@value": "Do the unglamorous work of passing the security review, signing the integration contract, and becoming the party with your name on outcomes. Permission precedes capability."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#howto_findUntrainableCorner",
    "@type": [
      "http://schema.org/HowTo"
    ],
    "http://schema.org/description": [
      {
        "@value": "A strategic playbook derived from the article's thesis for companies seeking durable AI value."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "How to Build in the Untrainable Corner"
      }
    ],
    "http://schema.org/step": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#step1_getInside"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#step2_arrangePrivateReality"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#step3_writeDownWhatGoodMeans"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#step4_priceOutcomes"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#step5_avoidCapitalWar"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#step6_keepSteppingForward"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#step6_keepSteppingForward",
    "@type": [
      "http://schema.org/HowToStep"
    ],
    "http://schema.org/name": [
      {
        "@value": "Keep stepping toward the not-yet-scored"
      }
    ],
    "http://schema.org/position": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#integer",
        "@value": 6
      }
    ],
    "http://schema.org/text": [
      {
        "@value": "The untrainable ground shrinks continuously. Don't find a defensible spot and rest. Re-underwrite constantly, moving toward whatever work cannot yet be benchmarked."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#evidence_mitStudy",
    "@type": [
      "http://schema.org/ScholarlyArticle"
    ],
    "http://schema.org/about": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#concept_illegibleCorrectness"
      }
    ],
    "http://schema.org/author": [
      {
        "@value": "Mert Demirer et al."
      }
    ],
    "http://schema.org/description": [
      {
        "@value": "Across more than 100,000 developers, the latest coding agents lifted code written by ~180% but code actually shipped by only ~30%. Writing got cheap; the rest still runs through a person."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "MIT Study on Coding Agent Productivity"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#gloss_moat",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "In competitive strategy, a durable barrier protecting a business from rivals. In the AI context, the untrainable corner argues that moats come from private data and trust relationships, not model quality."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Moat"
      }
    ],
    "http://www.w3.org/2002/07/owl#sameAs": [
      {
        "@id": "http://dbpedia.org/resource/Economic_moat"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq2",
    "@type": [
      "http://schema.org/Question"
    ],
    "http://schema.org/acceptedAnswer": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#faq2answer"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Why can't a smarter model win in markets like consumer chat?"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#step3_writeDownWhatGoodMeans",
    "@type": [
      "http://schema.org/HowToStep"
    ],
    "http://schema.org/name": [
      {
        "@value": "Write down what good means"
      }
    ],
    "http://schema.org/position": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#integer",
        "@value": 3
      }
    ],
    "http://schema.org/text": [
      {
        "@value": "Accumulate enough in-domain judgments to define evaluation criteria. Companies that own the definition of 'resolved' or author the legal benchmark earn standing no frontier lab can buy."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq4",
    "@type": [
      "http://schema.org/Question"
    ],
    "http://schema.org/acceptedAnswer": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#faq4answer"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "What is the 'absorption frontier' and why does it keep rising?"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#thesis_untrainableValue",
    "@type": [
      "http://schema.org/CreativeWork"
    ],
    "http://schema.org/description": [
      {
        "@value": "Durable AI value concentrates in work whose correctness is private, expensive to verify externally, and locked inside systems requiring trust and permission — the 'untrainable corner' no benchmark or general model can colonize."
      }
    ],
    "http://schema.org/genre": [
      {
        "@id": "http://dbpedia.org/resource/Thesis"
      }
    ],
    "http://schema.org/hasPart": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#claim_benchmarkEatsItself"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#claim_intelligenceNotBottleneck"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#claim_absorptionFrontierRises"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#claim_thinWrapperAbsorbed"
      },
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#claim_offenseHarder"
      }
    ],
    "http://schema.org/isPartOf": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#article"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "The Untrainable Thesis"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq1",
    "@type": [
      "http://schema.org/Question"
    ],
    "http://schema.org/acceptedAnswer": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#faq1answer"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Why is investor despair about AI investability only 'half right'?"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#concept_untrainableCorner",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "Frontier work whose correctness exists only in private data — the 2x2 quadrant where tasks are not saturated and answers are not publicly verifiable, making it resistant to general model commoditization."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "The Untrainable Corner"
      }
    ],
    "http://www.w3.org/2002/07/owl#sameAs": [
      {
        "@id": "http://dbpedia.org/resource/Proprietary_information"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq3",
    "@type": [
      "http://schema.org/Question"
    ],
    "http://schema.org/acceptedAnswer": [
      {
        "@id": "https://saranormous.substack.com/p/the-untrainable#faq3answer"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "What makes the law firm M&A example archetypal?"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#step2_arrangePrivateReality",
    "@type": [
      "http://schema.org/HowToStep"
    ],
    "http://schema.org/name": [
      {
        "@value": "Arrange the customer's private reality"
      }
    ],
    "http://schema.org/position": [
      {
        "@type": "http://www.w3.org/2001/XMLSchema#integer",
        "@value": 2
      }
    ],
    "http://schema.org/text": [
      {
        "@value": "Connect the model to the firm's private data, internal tools, and workflows. The translation of a company's private reality into model-actionable form is hard to copy and never ends."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq4answer",
    "@type": [
      "http://schema.org/Answer"
    ],
    "http://schema.org/text": [
      {
        "@value": "The absorption frontier is the expanding boundary at which model labs pull application scaffolding (retrieval, routing, tool use, reasoning policy) into base weights. It rises because more of the measurable work is continuously being benchmarked and then trained against, narrowing the app layer unless companies move into private, illegible territory."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#sarahGuo",
    "@type": [
      "http://schema.org/Person"
    ],
    "http://schema.org/description": [
      {
        "@value": "AI-focused venture investor and author of the Saranormous Substack."
      }
    ],
    "http://schema.org/jobTitle": [
      {
        "@value": "Venture Capital Investor"
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Sarah Guo"
      }
    ],
    "http://schema.org/sameAs": [
      {
        "@id": "https://substack.com/@saranormous"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://saranormous.substack.com/"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#gloss_inference",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "Running a trained model to produce outputs. While token price commoditizes on schedule, reliability, and guaranteed compute access during high traffic do not."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Inference (AI Serving)"
      }
    ],
    "http://www.w3.org/2002/07/owl#sameAs": [
      {
        "@id": "http://dbpedia.org/resource/Machine_learning#Inference"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_harvey",
    "@type": [
      "http://schema.org/Organization"
    ],
    "http://schema.org/description": [
      {
        "@value": "AI legal company that publishes a benchmark for law — earning the right to define what 'good' means in legal work through real adoption."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Harvey"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://www.harvey.ai/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq2answer",
    "@type": [
      "http://schema.org/Answer"
    ],
    "http://schema.org/text": [
      {
        "@value": "ChatGPT held its lead through years of real competition not on model quality alone. Gemini is gaining share via Android and Search, not a better model. Trust, habit, and distribution are non-model advantages the public weighs."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#faq1answer",
    "@type": [
      "http://schema.org/Answer"
    ],
    "http://schema.org/text": [
      {
        "@value": "The despair correctly identifies that thin wrappers over frontier models face absorption. It is wrong about what survives: private ground truth, domain trust, and illegible correctness create durable value the model layer cannot reach."
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_sierra",
    "@type": [
      "http://schema.org/Organization"
    ],
    "http://schema.org/description": [
      {
        "@value": "AI customer service company that charges only when its agent resolves a customer issue — outcome-based pricing that works because Sierra owns the definition of 'resolved'."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Sierra"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://sierra.ai/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_openEvidence",
    "@type": [
      "http://schema.org/SoftwareApplication"
    ],
    "http://schema.org/description": [
      {
        "@value": "AI medical reference platform used daily by a majority of American doctors."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "OpenEvidence"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://www.openevidence.com/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#claim_thinWrapperAbsorbed",
    "@type": [
      "http://schema.org/Statement"
    ],
    "http://schema.org/description": [
      {
        "@value": "Companies built merely on top of a frontier model with no proprietary data or trust integration are at severe absorption risk. The despair about AI investability is 'half right' on this point."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Thin wrappers are being absorbed"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#org_fireworks",
    "@type": [
      "http://schema.org/SoftwareApplication"
    ],
    "http://schema.org/description": [
      {
        "@value": "Inference serving provider alongside Baseten; reliability under real traffic and guaranteed access to scarce compute are non-commodity differentiators."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Fireworks AI"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://fireworks.ai/#this"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#concept_illegibleCorrectness",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "A form of truth that cannot be read off any leaderboard and is only discovered by running a complex system in the world over time. Decade-old codebases, hospital workflows, and legal matter flows exemplify this."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Illegible Correctness"
      }
    ],
    "http://www.w3.org/2002/07/owl#sameAs": [
      {
        "@id": "http://dbpedia.org/resource/Tacit_knowledge"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#concept_twoByTwo",
    "@type": [
      "http://schema.org/DefinedTerm"
    ],
    "http://schema.org/description": [
      {
        "@value": "A strategic framework: (1) Saturated + public answers = commodity tokens for open models; (2) Frontier + public answers = lab territory (free eval); (3) Saturated + private = partial moat; (4) Frontier + private = the untrainable corner, where durable value lives."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Saturated × Public/Private 2×2"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#claim_absorptionFrontierRises",
    "@type": [
      "http://schema.org/Statement"
    ],
    "http://schema.org/description": [
      {
        "@value": "Labs pull scaffolding (retrieval, routing, tool use, reasoning policy) into weights, narrowing the app layer. The untrainable ground shrinks under whoever stands on it; companies must continuously step toward what can't yet be scored."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "The absorption frontier keeps rising"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#claim_benchmarkEatsItself",
    "@type": [
      "http://schema.org/Statement"
    ],
    "http://schema.org/description": [
      {
        "@value": "A thing you can measure is a thing you can train against. Once work is checkable cheaply, it commoditizes toward open models at lowest cost. The measurable is always on its way to commodity."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Benchmarks are self-defeating"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#claim_offenseHarder",
    "@type": [
      "http://schema.org/Statement"
    ],
    "http://schema.org/description": [
      {
        "@value": "The model cannot tell you what is worth building. Identifying unaddressed use cases before the rest of the market is a uniquely human judgment — perhaps intent is an even scarcer input than compute."
      }
    ],
    "http://schema.org/name": [
      {
        "@value": "Choosing what to build is harder than defending what you have"
      }
    ]
  },
  {
    "@id": "https://saranormous.substack.com/p/the-untrainable#saranormousSubstack",
    "@type": [
      "http://schema.org/Organization"
    ],
    "http://schema.org/name": [
      {
        "@value": "Saranormous (Substack)"
      }
    ],
    "http://schema.org/url": [
      {
        "@id": "https://saranormous.substack.com/"
      }
    ]
  }
]