{
  "schemaVersion": 1,
  "id": "crg-system",
  "title": "The commitment layer for AI-native chip design",
  "category": "company",
  "format": "System release",
  "description": "CRG Systems v1.5.0 connects formal theory, reproducible validation, independent replay, and an agent-native boundary in one cumulative system.",
  "canonicalUrl": "https://semiaifoundry.com/updates/the-commitment-layer-for-ai-native-chip-design/",
  "publishedDate": "2026-09-03",
  "revisionDate": "2026-10-07",
  "author": "F. M. Cinnor",
  "topic": "Computation–Realization Geometry",
  "series": null,
  "pursuitIds": [
    "computation-and-realization"
  ],
  "license": null,
  "rightsUrl": "https://semiaifoundry.com/terms-of-use/",
  "version": "1.5.0",
  "status": "Published public page",
  "evidenceScope": null,
  "contentCoverage": "published-page",
  "artifactUrl": null,
  "interactiveUrl": null,
  "related": [
    {
      "id": "crg-theory",
      "relationship": "Theoretical foundations",
      "description": "The six-manuscript program establishes realization geometry, safe commitment, scaling, and architecture selection.",
      "title": "A Communication Theory of Computation Realization",
      "canonicalUrl": "https://semiaifoundry.com/research/computation-realization-geometry/"
    },
    {
      "id": "agentic-brain",
      "relationship": "Agent authority",
      "description": "A separate framework for evidence-gated authority and runtime governance in consequential agent systems.",
      "title": "The Agentic Brain",
      "canonicalUrl": "https://semiaifoundry.com/research/the-agentic-brain/"
    }
  ],
  "supersedes": [],
  "supersededBy": [],
  "contentHash": "e6052b34d4b4405365f53ca357b206ee3791bd9212ae54137eba5f33cd50a711",
  "contentLength": 14480,
  "jsonUrl": "https://semiaifoundry.com/agents/publications/crg-system.json",
  "markdownUrl": "https://semiaifoundry.com/agents/publications/crg-system.md",
  "sections": [
    {
      "id": "section-1",
      "title": "The commitment layer for AI-native chip design.",
      "start": 70,
      "end": 1275
    },
    {
      "id": "section-2",
      "title": "The bottleneck has moved.",
      "start": 1275,
      "end": 2252
    },
    {
      "id": "section-3",
      "title": "The layer between possibility and commitment.",
      "start": 2252,
      "end": 3134
    },
    {
      "id": "section-4",
      "title": "From a formal law to an agent-native system.",
      "start": 3134,
      "end": 4479
    },
    {
      "id": "section-5",
      "title": "The cost of forgetting is measurable.",
      "start": 4479,
      "end": 4776
    },
    {
      "id": "section-6",
      "title": "Build the policy without seeing the test.",
      "start": 4776,
      "end": 5015
    },
    {
      "id": "section-7",
      "title": "Ask whether the retained set survives new conditions.",
      "start": 5015,
      "end": 5301
    },
    {
      "id": "section-8",
      "title": "Change what is legally possible.",
      "start": 5301,
      "end": 5913
    },
    {
      "id": "section-9",
      "title": "Held-out conditions",
      "start": 5913,
      "end": 6005
    },
    {
      "id": "section-10",
      "title": "Technology shift",
      "start": 6005,
      "end": 7617
    },
    {
      "id": "section-11",
      "title": "Six releases. One commitment stack.",
      "start": 7617,
      "end": 8963
    },
    {
      "id": "section-12",
      "title": "An integrated system, not a thin wrapper.",
      "start": 8963,
      "end": 9799
    },
    {
      "id": "section-13",
      "title": "The depth of the cumulative platform is measurable in years, not feature sprints.",
      "start": 9799,
      "end": 10585
    },
    {
      "id": "section-14",
      "title": "A decision travels with its proof.",
      "start": 10585,
      "end": 11530
    },
    {
      "id": "section-15",
      "title": "May this distinction be discarded?",
      "start": 11530,
      "end": 12520
    },
    {
      "id": "section-16",
      "title": "Independently replayable by design.",
      "start": 12520,
      "end": 13042
    },
    {
      "id": "section-17",
      "title": "Scale begins with one commitment boundary.",
      "start": 13042,
      "end": 14480
    }
  ],
  "citations": [
    {
      "title": "Explore v1.5.0 ↗",
      "url": "https://github.com/SemiAIFoundry/crg-systems/releases/tag/v1.5.0"
    },
    {
      "title": "Read the evidence ↓",
      "url": "https://semiaifoundry.com/updates/the-commitment-layer-for-ai-native-chip-design/#evidence"
    },
    {
      "title": "All updates",
      "url": "https://semiaifoundry.com/updates/"
    },
    {
      "title": "CRG research",
      "url": "https://semiaifoundry.com/research/computation-realization-geometry/"
    },
    {
      "title": "Read the frozen protocol ↗",
      "url": "https://github.com/SemiAIFoundry/computation-realization-geometry/blob/bf3964f300f3142470f40e0ff9b8d0548e17dd97/validation/retention/PREREGISTRATION.json"
    },
    {
      "title": "Inspect the retained results ↗",
      "url": "https://github.com/SemiAIFoundry/computation-realization-geometry/blob/bf3964f300f3142470f40e0ff9b8d0548e17dd97/validation/retention/expected/experiment_summary.json"
    },
    {
      "title": "Download v1.5.0 ↗",
      "url": "https://github.com/SemiAIFoundry/crg-systems/releases/download/v1.5.0/crg-systems-v1.5.0.tar.gz"
    },
    {
      "title": "Inspect the v1.5.0 distribution ↗",
      "url": "https://github.com/SemiAIFoundry/crg-systems/tree/v1.5.0"
    },
    {
      "title": "Discuss an integration →",
      "url": "https://semiaifoundry.com/contact/"
    },
    {
      "title": "Verify and install v1.5.0 ↗",
      "url": "https://github.com/SemiAIFoundry/crg-systems/blob/v1.5.0/README.md#user-content-install"
    }
  ],
  "recordHash": "56ca01902497c27966f400dd9816bddc1c7ac22d63873211b20fadae9002968f",
  "contentFormat": "text/markdown",
  "content": "September 3, 2026CRG Systems v1.5.0By F. M. Cinnor\n\nSystem milestone\n\n# The commitment layer for AI-native chip design.\n\nGeneration is becoming abundant. Elimination is becoming more consequential. CRG Systems makes design-space reduction explicit, scoped, and independently replayable, so autonomous design systems can explore broadly without silently erasing choices future decisions may still need.\n\n[Explore v1.5.0 ↗](https://github.com/SemiAIFoundry/crg-systems/releases/tag/v1.5.0) [Read the evidence ↓](https://semiaifoundry.com/updates/the-commitment-layer-for-ai-native-chip-design/#evidence)\n\nFrom broad design exploration to governed commitment A large connected field of design alternatives passes through the CRG commitment boundary. A retained family continues with a replayable record while evidence and change can reopen the decision.\n\nExplore broadlyCommit within scopeReopen when reality changes\n\nThe defining systems question\n\nWhat is the workflow justified in forgetting?\n\nOrchestration decides what runs.\n\nEDA determines what passes.\n\nCommitment infrastructure makes narrowing explicit, scoped, and replayable.\n\nWhy now[All updates](https://semiaifoundry.com/updates/)[CRG research](https://semiaifoundry.com/research/computation-realization-geometry/)\n\n## The bottleneck has moved.\n\nAI-native chip design is entering an age of abundance. Agents can generate more RTL variants, mappings, schedules, verification strategies, floorplans, package configurations, and repair paths than any engineering team can manually inspect.\n\nEvery autonomous design loop eventually narrows. Candidates are merged. Dominated points are pruned. Search stops. An architecture freezes. At that moment, the workflow makes a consequential claim, often silently: the distinctions it erased will not be needed by the decisions still ahead.\n\nThat claim needs infrastructure.\n\n> Verification asks whether a chosen design passes. Commitment asks whether choosing it destroyed information the next decision will need.\n\nCRG Systems provides that missing layer. It gives AI-native workflows _Realization Memory_: a durable account of the alternatives, identities, evidence, decisions, and reopening conditions behind consequential reductions.\n\nSystem position\n\n## The layer between possibility and commitment.\n\nCRG sits beneath agents and alongside existing engineering tools. It does not replace generation, EDA, simulation, signoff, or organizational authority. It makes the narrowing between them inspectable.\n\n01 / Explore**Models and agents**\n\nGenerative modelsDomain generatorsSearch and optimizationMulti-agent systems\n\n02 / Govern**CRG commitment layer**\n\nRealization identityDecision scopeCertificate, witness, or refusalReopening conditions\n\n03 / Realize**EDA and physical workflows**\n\nEvaluation and simulationVerification and signoffFoundry and packagingDeployment qualification\n\nEvidence and change\n\nReprice, revalidate, reopen, or regenerate as the operating context changes.\n\nAgents expand the search. CRG governs the commitments. Qualified engineering systems determine what becomes physically realizable.\n\nCapstone milestone\n\n## From a formal law to an agent-native system.\n\nCRG Systems v1.5.0 closes the initial theory-to-platform arc. Each stage is public, inspectable, and connected to the next.\n\n1.  01\n    \n    Formal theory\n    \n    ### Define what a safe reduction must preserve.\n    \n    A six-manuscript technical corpus formalizes computations, realization families, decision-visible geometry, and the Safe-Commitment Law.\n    \n    Decision-preservation laws\n2.  02\n    \n    Reproducible validation\n    \n    ### Test what breaks when narrowing happens too early.\n    \n    Frozen studies measure winner misses, regret under change, and the difference between repricing yesterday's survivors and reopening the family.\n    \n    Public protocols and results\n3.  03\n    \n    Operational system\n    \n    ### Carry exact semantics into deployable software.\n    \n    The cumulative 1.x runtime binds identity, retention, evidence, change, continued validity, portable records, and independent verification.\n    \n    Replayable runtime artifacts\n4.  04\n    \n    Agent-native boundary\n    \n    ### Put justified commitment in the agent's critical path.\n    \n    v1.5 adds causal receipts, bounded coordination, canonical package resolution, execution contracts, generated clients, and an Embedded Developer Kit preview.\n    \n    Integration-ready interfaces\n\nValidation\n\n## The cost of forgetting is measurable.\n\nThe validation program asks the question optimizers usually leave behind: what must survive reduction so that future decisions remain sound?\n\n**26,624**retention scenarios\n\n**4**Transformer workloads\n\n**1,024**capability-change scenarios\n\n01 / Construct\n\n### Build the policy without seeing the test.\n\nFive legal, nonvertical realizations of a Transformer microblock were evaluated across four workloads. Each policy could learn only from 512 construction scenarios per workload.\n\n02 / Stress\n\n### Ask whether the retained set survives new conditions.\n\nDisjoint seeds generated held-out, technology-shift, objective-shift, joint-shift, and controlled-perturbation cases. The headline comparison uses 1,024 held-out and 1,024 technology-shift scenarios per workload.\n\n03 / Reopen\n\n### Change what is legally possible.\n\nA separate 1,024-scenario track introduced a newly legal vertical realization, then compared rescoring yesterday's survivors with reopening and rebuilding the realization family.\n\nWinner miss\n\nThe later best legal realization was absent from the retained set.\n\nRelative regret\n\nThe normalized loss between the best retained choice and the true later best.\n\nMatched budget\n\nThe exact minimax baseline and CRG ledger cover retained the same two-to-four candidates on all four workloads.\n\nLater-winner miss rate\n\nHow often the retained set failed to include the later winner.\n\n### Held-out conditions\n\nInitial winner only**48.4%**\n\nCRG / minimax, matched budget**0%**\n\n### Technology shift\n\nInitial winner only**59.7%**\n\nCRG / minimax, matched budget**11.0%**\n\nCapability change\n\nMean relative regret after the legal realization family changed.\n\n**100%** change detection**97.1%** newly legal realization wins\n\nReprice retained options only**27.5%**\n\nReopen the realization family**0% observed**\n\n> No rescoring of yesterday's survivors can recover options that were never carried forward.\n\nWhat the results say Keeping today's winner is brittle. A decision-scoped set preserved every later winner under held-out conditions at the same two-to-four-candidate budget, while a technology shift exposed the limit of that construction scope. Universal-monotone retention kept all five nondominated candidates and recorded no misses in its tested unchanged-family scope. When a new capability changed the legal family, reopening restored zero observed regret; repricing the old survivors could not.\n\nWhat they do not say This is exact, finite-registry validation of retention and reopening semantics. It is not silicon PPA signoff, physical-implementation qualification, or a universal result across all chip designs. Construction and evaluation ledgers used disjoint, frozen seeds.\n\n[Read the frozen protocol ↗](https://github.com/SemiAIFoundry/computation-realization-geometry/blob/bf3964f300f3142470f40e0ff9b8d0548e17dd97/validation/retention/PREREGISTRATION.json) [Inspect the retained results ↗](https://github.com/SemiAIFoundry/computation-realization-geometry/blob/bf3964f300f3142470f40e0ff9b8d0548e17dd97/validation/retention/expected/experiment_summary.json)\n\nCumulative build\n\n## Six releases. One commitment stack.\n\nv1.5 is consequential because it completes a connected architecture rather than adding an isolated agent feature.\n\n1.  v1.0 / v1.0.1\n    \n    ### Exact foundation\n    \n    Identity, numerics, realization generation, geometry, certification, change, portable bundles, durable audit, and separate verification.\n    \n2.  v1.1.0\n    \n    ### Safe commitment\n    \n    Registry intake, decision-scoped retention, explicit pruning conditions, bounded-regret evidence, and portable records.\n    \n3.  v1.2.0\n    \n    ### Challenge and change\n    \n    Phase and stability analysis, certified deltas, comparison, and Decision Challenge.\n    \n4.  v1.3.0\n    \n    ### Qualified evidence\n    \n    Evidence applicability, uncertainty, qualification, selective revalidation, and controlled replacement.\n    \n5.  v1.4.0\n    \n    ### Durable validity\n    \n    Continued validity, lifecycle authority, reopening, recovery, migration evidence, and resilience.\n    \n6.  v1.5.0\n    \n    ### Agent-native boundary\n    \n    Design-space reduction, causal receipts, bounded coordination, package resolution, restricted execution, and end-to-end agent surfaces.\n    \n\nExact identity → governed retention → challengeable change → qualified evidence → durable validity and authority → agent-native realization memory\n\nPlatform depth\n\n## An integrated system, not a thin wrapper.\n\nThe v1.5 distribution exposes one canonical semantic system across product, integration, storage, lifecycle, and verification surfaces.\n\n**17**component packages\n\n**100**public schemas\n\n**106**API operations\n\n**83**executable CLI leaves\n\n**17**Studio workspaces\n\n**7**reference packs\n\nReference compatibility**47 / 47**\n\nmandatory cases passed across 13 declared profiles\n\nIndependent implementation**10 / 10**\n\napplicable cases passed by the zero-dependency Node.js verifier\n\nThe published status is _compatible, not officially conformant_: every mandatory case passes, while no external governance authority is represented as having issued a designation. [Read the v1.5.0 release record ↗](https://github.com/SemiAIFoundry/crg-systems/releases/tag/v1.5.0)\n\nConventional build equivalent\n\n### The depth of the cumulative platform is measurable in years, not feature sprints.\n\nAn internal greenfield engineering-equivalence assessment estimates the conventional effort required to reproduce a comparable concept-to-v1.5 platform, including theory-to-system translation, exact computation, product surfaces, security, release discipline, and independent verification.\n\n**30–42**FTE-years\n\n**30–42**calendar months\n\n**$10–18M**estimated 2026 U.S. build effort\n\nEngineering-equivalence planning estimate, not historical expenditure, audited cost, product pricing, company valuation, customer adoption, or realized return. The range reflects a technical-baseline, work-breakdown, and sensitivity method using U.S. labor and employer-compensation benchmarks.\n\nv1.5 agent boundary\n\n## A decision travels with its proof.\n\nA proposed reduction becomes a bounded, portable statement carrying the derived realization family, declared decision scope, retained and pruned realizations, resource limits, evidence, and an independent reconstruction path.\n\nAuthority boundary\n\n**Assessment is not authorization.** A CRG result does not authorize execution, publication, tape-out, or repository mutation. Accountable organizational policy remains in control.\n\nDeclared request\n\nAn agent or orchestrator supplies the problem, not the conclusion.\n\n-   **Operator graph**\n    \n    The computation and legal transformations.\n    \n-   **Explicit bounds**\n    \n    Exact inputs, resource limits, and completeness basis.\n    \n-   **Decision scope**\n    \n    The future observations the reduction must preserve.\n    \n-   **Qualified evidence**\n    \n    EDA, simulation, and source-adapter results with provenance.\n    \n\nCRG commitment boundary\n\n### May this distinction be discarded?\n\n1.  01\n    \n    Derive the finite closure, legal realization family, and exact geometry.\n    \n2.  02\n    \n    Test the proposed reduction against the declared future-decision scope.\n    \n3.  03\n    \n    Bind the result to evidence, resource limits, and reopening conditions.\n    \n\nTyped outcome\n\nUncertainty never collapses into an unsupported success.\n\n-   Certificate\n    \n    The reduction is justified within its declared scope.\n    \n-   Regret witness\n    \n    A future decision can distinguish an option that would be pruned.\n    \n-   Typed refusal\n    \n    Evidence, completeness, support, or resource limits are insufficient.\n    \n\nChange routing\n\nThe return path depends on what changed.\n\n**Prices or weights**Reprice\n\n**Evidence validity**Revalidate\n\n**Legality or capability**Reopen\n\n**Generation boundary**Regenerate\n\nCRG governs the permission to forget. It does not choose what to build or who may execute.\n\nPortable commitment record\n\n### Independently replayable by design.\n\nThe verifier reconstructs supported claims rather than trusting producer status.\n\nRegistered family\n\nIdentity and completeness basis\n\nDecision scope\n\nFuture observations to preserve\n\nReduction\n\nRetained and pruned realizations\n\nResult\n\nCertificate, witness, or refusal\n\nValidity\n\nEvidence dependencies and reopening conditions\n\nVerification\n\nCode-separated reconstruction path\n\n> Build your differentiation above CRG. Do not rebuild the trust substrate beneath it.\n\nAdoption path\n\n## Scale begins with one commitment boundary.\n\nThe initial 1.x foundation is complete and built for scale. v1.5.0 is ready for bounded integration into AI-native design workflows: begin where an agent or optimization process can erase a choice, declare what future decisions the reduction must preserve, and independently replay the result.\n\nExpand from that tested boundary as the organization qualifies its tools, policies, evidence, and deployment environment.\n\n01\n\nChoose one consequential reduction.\n\n02\n\nInspect the certificate, witness, or refusal.\n\n03\n\nReplay independently, then change an assumption.\n\n[Download v1.5.0 ↗](https://github.com/SemiAIFoundry/crg-systems/releases/download/v1.5.0/crg-systems-v1.5.0.tar.gz) [Inspect the v1.5.0 distribution ↗](https://github.com/SemiAIFoundry/crg-systems/tree/v1.5.0) [Discuss an integration →](https://semiaifoundry.com/contact/)\n\nCRG Systems is source-available for permitted research, educational, and other noncommercial use with attribution. Commercial use requires a separate written commercial license from Semi AI Foundry, LLC. The release license controls.\n\nExplore freely.  \nEliminate with reasons.  \nCommit within scope.  \nReopen when reality changes.\n\n[Explore the research program →](https://semiaifoundry.com/research/computation-realization-geometry/)[Verify and install v1.5.0 ↗](https://github.com/SemiAIFoundry/crg-systems/blob/v1.5.0/README.md#user-content-install)"
}