{
  "section": "theory",
  "kind": "theory",
  "canonical": "https://evemisslab.com/ai/theory/",
  "count": 17,
  "records": [
    {
      "id": "THY-2026-0005",
      "kind": "theory",
      "label": "PACC conjecture — the four-level convergence ladder",
      "created_at": "2026-09-08",
      "updated_at": "2026-09-09",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E3",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0005/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "PACC-B (behavioural: D_B ≤ ε), PACC-R (representational: a low-complexity map Φ: S_N → S_P valid on held-out tasks, K(Φ) ≤ κ), PACC-D (dynamical: Φ∘U_N ≈ U_P∘Φ) and PACC-A (architecture attractor across ≥3 independent designs). The conjecture pre-registers its own failure modes F1–F6 — behavioural divergence, no low-complexity map, update diagram fails, intervention divergence, architecture diverges under scale, probability-specific advantage persists — so that 'similar' and 'probability' cannot be redefined after the fact.",
        "eml_summary_zh": "PACC-B（行為：D_B ≤ ε）、PACC-R（表徵：在 held-out 任務上仍有效的低複雜度映射 Φ: S_N → S_P，K(Φ) ≤ κ）、PACC-D（動力學：Φ∘U_N ≈ U_P∘Φ）與 PACC-A（≥3 個獨立設計的架構吸引子）。猜想預先登記自己的失敗模式 F1–F6——行為分歧、無低複雜度映射、更新圖失敗、干預分歧、規模下架構分歧、概率特有優勢持續——讓「相似」與「概率」不能事後重新定義。",
        "eml_label_zh": "PACC 猜想——四層收斂階梯",
        "eml_primary_domain": "Model Representation",
        "eml_domains": [
          "Formal AI",
          "Reasoning"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "Primitive-level non-probabilistic system: canonical state and required updates do not need distributions, Kolmogorov semantics, Bayesian posteriors, a probability-simplex container or sampling; raw weights, orders, signed relations, tensions and constraints are allowed.",
          "Probability-like representation: a low-complexity map to the simplex that roughly preserves ranking, choice, relative strength and evidence response."
        ],
        "eml_assumptions": [
          "Intelligence constraints: partial observation, competing alternatives, finite resources, repeated update, action selection, changing environment."
        ],
        "eml_claims": [
          "Probability may emerge from competition without being the original semantics of competition.",
          "AI's feasible computational forms may be far fewer than the theories that describe them."
        ],
        "eml_predictions": [
          "Outcome I behavioural equivalence only → multiple realizability; II representational duality; III dynamical equivalence → alternative coordinates of a common computation; IV computational attractor."
        ],
        "eml_falsification_conditions": [
          "F1 D_B ≫ ε_B; F2 no generalizing low-complexity Φ; F3 D_U large; F4 intervention divergence; F5 D_A rises with scale; F6 probability-specific advantage from the update itself."
        ],
        "eml_known_limitations": [
          "PACC-A cannot pass with two independent families; the strongest current statement is a nontrivial convergence basin with an atlas of partially overlapping probability-like charts, not a universal attractor.",
          "A conjecture, tested so far only on synthetic data with theoretical reasoning: theoretically possible is not actually possible."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0005/",
      "json": "/ai/theory/THY-2026-0005/index.json"
    },
    {
      "id": "THY-2026-0001",
      "kind": "theory",
      "label": "Asymmetric spacetime tension: temporally heterogeneous world knowledge",
      "created_at": "2026-09-07",
      "updated_at": "2026-09-08",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E2",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0001/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Asymmetry is lifted from edge direction or weight to the effective time scale of nodes and relations. Each node carries stability, information-decay rate, update tension, system impact and local valid time; refresh is triggered by tension, external disturbance, information age, change velocity and event relevance rather than by one global clock, so stable knowledge sleeps, volatile knowledge refreshes often, and a rarely changing high-impact node triggers wide dependency recomputation when it does change.",
        "eml_summary_zh": "把非對稱性從邊的方向或權重，提升到節點與關係的有效時間尺度。每個節點帶有穩定性、資訊衰減率、更新張力、系統影響度與局部有效時間；刷新由張力、外部擾動、資訊年齡、變化速度與事件相關性觸發，而不是一個全域時鐘——穩定知識沉睡、快變知識高頻更新、低頻高影響的基礎節點一旦改變就觸發大範圍依賴重算。",
        "eml_label_zh": "非對稱時空張力：時間異質的世界知識",
        "eml_primary_domain": "Context & Memory",
        "eml_domains": [
          "Computation",
          "AI Architecture"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_assumptions": [
          "Knowledge stability is highly non-uniform across formal, physical, social and real-time domains.",
          "Update cost is not free; over-refresh and under-refresh are both failures."
        ],
        "eml_claims": [
          "World knowledge is temporally heterogeneous.",
          "Freshness is not a timestamp: it is decided jointly by world change rate, source reliability, dependency structure, task risk and system impact (Paper 02)."
        ],
        "eml_predictions": [
          "Architecture contribution AC(λ=0) ≈ 0 in a static world, while AC(λ_heterogeneous) > 0; a fixed synchronous-refresh baseline should be worse in highly heterogeneous time-scale environments (Paper 11 §80–82)."
        ],
        "eml_falsification_conditions": [
          "No measurable staleness/recomputation advantage over fixed-TTL and event-driven refresh under heterogeneous dynamics.",
          "The tension field is only an LLM guess with no runtime enforcement — then the theory has not been tested (Paper 11 §29)."
        ],
        "eml_known_limitations": [
          "R1 found that in an event-rich deterministic world, pure event invalidation is cheaper than tension scheduling (AER 3 recomputations vs 2)."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0001/",
      "json": "/ai/theory/THY-2026-0001/index.json"
    },
    {
      "id": "THY-2026-0002",
      "kind": "theory",
      "label": "Canonical symbolic state and candidate → verify → commit authority",
      "created_at": "2026-09-07",
      "updated_at": "2026-09-08",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E2",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0002/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Natural language is input and output, never the canonical state. Text is parsed, normalized, semantically bound and source-tagged into comparable, verifiable symbolic structure; output is rendered from that state. Model and tool outputs are candidate evidence, and only a committer, after a passing verifier and an expected-version check, may mutate canonical state.",
        "eml_summary_zh": "自然語言是輸入與輸出，永遠不是 canonical state。文字經解析、正規化、語義綁定與來源標記，轉成可比較、可驗證的符號結構；輸出從該狀態 render 出來。模型與工具的輸出是候選證據，只有 committer 在 verifier 通過與 expected-version 檢查後才能改動 canonical state。",
        "eml_label_zh": "Canonical 符號狀態與 candidate → verify → commit 權限",
        "eml_primary_domain": "AI Architecture",
        "eml_domains": [
          "AI-native Systems"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_assumptions": [
          "Ambiguity, synonymy, context dependence and self-contamination make language a poor sole carrier of state."
        ],
        "eml_claims": [
          "Text, tables, code, graphs, numbers and actions are different output projections of one canonical state.",
          "Inference/tool output → Candidate → Verify → Commit, never direct state mutation."
        ],
        "eml_predictions": [
          "A no-provenance fact candidate is rejected before commit; two valid candidates on the same base version yield exactly one commit and one VersionConflict."
        ],
        "eml_falsification_conditions": [
          "A strengthened baseline with direct mutable memory reproduces the same state-integrity outcomes at lower cost."
        ],
        "eml_known_limitations": [
          "A centralized application-level gate reproduces the tested semantics (R3); the remaining difference is where the obligation lives, not what can be computed."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0002/",
      "json": "/ai/theory/THY-2026-0002/index.json"
    },
    {
      "id": "THY-2026-0003",
      "kind": "theory",
      "label": "Intelligent architecture attractor",
      "created_at": "2026-09-08",
      "updated_at": "2026-09-08",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E2",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0003/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Different first principles may compile into the same small family of computational forms. The comparison framework separates six levels of difference — code, primitive, computation trace, architectural state semantics, observable behaviour, resource efficiency — and asks who owns state, what may rewrite canonical state, and whether dynamics are preserved under low-cost mapping. The series ends with a fixed verdict map: Distinct Advantage, Operational Convergence, Behavioral Equivalence Only, Inconclusive, Architecture Worse.",
        "eml_summary_zh": "不同的第一原理可能編譯成同一小族計算形態。比較框架區分六個差異層級——程式碼、primitive、計算軌跡、架構狀態語義、可觀測行為、資源效率——並追問誰持有狀態、什麼可以改寫 canonical state、動態結構在低成本映射下是否保持。系列以固定的判決表結束：Distinct Advantage、Operational Convergence、Behavioral Equivalence Only、Inconclusive、Architecture Worse。",
        "eml_label_zh": "智能架構吸引子",
        "eml_primary_domain": "AI Architecture",
        "eml_domains": [
          "Evaluation",
          "Formal AI"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_assumptions": [
          "The executed architecture is the architecture (runtime-truth principle); spec cannot defend runtime."
        ],
        "eml_claims": [
          "Theory difference matters only insofar as it survives into measurable state, computation, behaviour or resource use; if it does not survive, the convergence itself becomes the phenomenon to explain.",
          "If every possible outcome is interpreted as confirmation, the theory has explained nothing."
        ],
        "eml_predictions": [
          "Independent architectures' ablation cores intersect (an attractor core) if the attractor is real."
        ],
        "eml_falsification_conditions": [
          "Distinct Advantage: stable multidimensional advantage over baselines across models, seeds, tasks and horizons weakens the convergence suspicion.",
          "Architecture Worse: added structure is operationally unnecessary or harmful."
        ],
        "eml_known_limitations": [
          "One implementation provides at most candidate evidence for convergence; an attractor study needs n ≫ 2 independent starting points (Paper 11 §53)."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0003/",
      "json": "/ai/theory/THY-2026-0003/index.json"
    },
    {
      "id": "THY-2026-0004",
      "kind": "theory",
      "label": "Probability is not Bayesian; Bayes cannot self-authorize its premises",
      "created_at": "2026-09-10",
      "updated_at": "2026-09-08",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0004/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "A system can be stochastic, probabilistic or probability-shaped without performing Bayesian conditionalization, and can look Bayesian without a real prior, likelihood or posterior. A layered vocabulary (stochastic, probabilistic, Bayesian-like, exact, approximate, generalized Bayesian) and a Bayesian authenticity test check whether an update is substantively Bayesian or merely redescribed as such; and the update rule itself — prior, likelihood, hypothesis space — needs a justification that Bayes' rule does not supply (Papers 09–10).",
        "eml_summary_zh": "一個系統可以是隨機的、概率的或概率形狀的，卻沒有做 Bayesian conditionalization；也可以看起來像 Bayesian，卻沒有真正的 prior、likelihood 或 posterior。一套分層詞彙（stochastic、probabilistic、Bayesian-like、exact、approximate、generalized Bayesian）與「Bayesian authenticity test」檢查一次更新是實質 Bayesian 還是事後被重新描述成 Bayesian；而更新規則本身——prior、likelihood、假設空間——需要一個 Bayes 公式給不出的授權（第 9–10 篇）。",
        "eml_label_zh": "概率不等於貝葉斯；貝葉斯不能自我授權前提",
        "eml_primary_domain": "Formal AI",
        "eml_domains": [
          "Reasoning"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_claims": [
          "'This is a probabilistic system' and 'this is a Bayesian system' are not equivalent statements.",
          "Bayesian updating is one belief-revision operator among several; the epistemic router should choose the operator explicitly and record its provenance."
        ],
        "eml_predictions": [
          "An adaptive epistemic router beats AlwaysBayes, AlwaysLogic and AlwaysRobust across mixed domains only if operator choice is explicit and traced (Paper 11 §86)."
        ],
        "eml_falsification_conditions": [
          "Fixed Bayesian updating dominates every mixed-domain test at equal cost."
        ],
        "eml_known_limitations": [
          "Formal/conceptual so far; the AER-0 router implements four operators but no mixed-domain routing benchmark has been run."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0004/",
      "json": "/ai/theory/THY-2026-0004/index.json"
    },
    {
      "id": "THY-2026-0006",
      "kind": "theory",
      "label": "AER-ECT: a mandatory epistemic commit transaction boundary",
      "created_at": "2026-09-08",
      "updated_at": "2026-09-08",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E2",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0006/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Every canonical knowledge mutation must pass one mandatory epistemic transaction boundary binding claim, evidence, provenance, valid and observed time, expected version, epistemic operator, verification policy, authority, dependencies and refresh policy: Propose → Verify → Authorize → CompareAndCommit → BindDependencies → Audit. Its parts are not new (truth maintenance, transaction logic, PROV, bitemporal state, PDP/PEP); the hypothesis is that the tuple is mandatory at the canonical write boundary — an epistemic reference-monitor-like boundary, not a proven tamperproof monitor.",
        "eml_summary_zh": "每一次 canonical knowledge 的改動都必須通過一個強制的認識論交易邊界，綁定 claim、evidence、provenance、有效時間與觀察時間、expected version、認識論算子、驗證政策、權限、依賴與 refresh policy：Propose → Verify → Authorize → CompareAndCommit → BindDependencies → Audit。零件都不新（truth maintenance、transaction logic、PROV、bitemporal state、PDP/PEP）；假說是這個 tuple 在 canonical write 邊界上是強制的——一個類 reference-monitor 的認識論邊界，而非已證明 tamperproof 的 monitor。",
        "eml_label_zh": "AER-ECT：強制的認識論 commit 交易邊界",
        "eml_primary_domain": "AI Infrastructure",
        "eml_domains": [
          "AI Architecture",
          "Agent Systems"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_claims": [
          "Architectural elevation = same semantic capability + smaller policy-scattering surface + mandatory canonical boundary; not computational uniqueness.",
          "When accepted knowledge has operational consequences, epistemic governance may deserve the architectural status that transaction boundaries have in databases."
        ],
        "eml_predictions": [
          "Policy sites and rule placements scale O(NR) when scattered versus O(R) under one gate; a new caller does not require new policy placement.",
          "Internal consistency ≠ authenticity: without an external trust anchor a coherent full-DB forgery passes internal audit."
        ],
        "eml_falsification_conditions": [
          "Bypass is easy under accidental or adversarial pressure — then ECT is merely a convenient API.",
          "Hostile same-process code or a full DB writer defeats local controls (accepted: tamperproofness is not claimed)."
        ],
        "eml_known_limitations": [
          "Primitive, transaction, provenance, versioning, belief-revision and temporal novelty are weak or not claimed (R3 novelty verdict); OS privilege isolation, signer compromise and cross-resource ACID are not measured (R6)."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0006/",
      "json": "/ai/theory/THY-2026-0006/index.json"
    },
    {
      "id": "THY-2026-0007",
      "kind": "theory",
      "label": "Capability memory and substrate-neutral compute: Reuse ≻ Adapt ≻ Create",
      "created_at": "2026-09-08",
      "updated_at": "2026-09-08",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E2",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0007/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_aes/extract.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Beyond facts, the system keeps the algorithms, tools, execution contracts, costs, versions, applicability conditions and success/failure history it has used, and prefers reusing a known solution path over adapting one over creating one. Algorithms and compute containers are different layers: any environment that accepts representable input, performs a valid state transition and returns readable output is a container with its own cost, latency, error and availability model, so algorithm/container pairs are selected jointly.",
        "eml_summary_zh": "除了事實之外，系統也保存用過的算法、工具、執行契約、成本、版本、適用條件與成敗歷史，並且偏好重用已知求解路徑，勝過調整，再勝過重造。算法與計算容器是不同層：任何能接受可表示輸入、執行有效狀態轉換並回傳可讀輸出的環境都是一個容器，各有自己的成本、延遲、誤差與可用性模型，因此算法／容器成對聯合選擇。",
        "eml_label_zh": "能力記憶與載體中立計算：Reuse ≻ Adapt ≻ Create",
        "eml_primary_domain": "Computation",
        "eml_domains": [
          "AI Architecture",
          "AI Infrastructure"
        ],
        "eml_program_id": "PRG-2026-0001",
        "eml_data_basis": "THEORY",
        "eml_claims": [
          "Reuse ≻ Adapt ≻ Create.",
          "Algorithm ≠ container; a capability/container pair is the unit of selection and of execution trace."
        ],
        "eml_predictions": [
          "Planning cost falls with repeated related tasks; ReuseGain grows with task similarity within the applicability range, and constraint mismatch produces measurable false reuse (Paper 11 §83–84)."
        ],
        "eml_falsification_conditions": [
          "No planning-cost reduction with experience; negative transfer dominates."
        ],
        "eml_known_limitations": [
          "R1/R2: persistent workflow memory outside the model is not unique to AER; the capability/container separation was 'AER default distinct' only because no core LangGraph primitive was found, not because it cannot be built there."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Sol (GPT-5.6, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0007/",
      "json": "/ai/theory/THY-2026-0007/index.json"
    },
    {
      "id": "THY-2026-0101",
      "kind": "theory",
      "label": "Turn decomposition and externally loopless intelligence",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0101/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "A 'turn' is decomposed into the vector T = (U user turns, G generation trajectories, I model invocations, L external feedback loops, R retries, S selection/verification, P physical execution trace). The cleanest single pass is U=1, G=1, R=1, L=0, S=0, and it excludes only action→new-evidence→replanning, not autoregressive sequential computation. Loops come in five kinds (tool, environment, verifier, retry, candidate/selection), capability lives in four layers (intrinsic, elicited, system, product), and every 'one-turn' claim projects onto the event vector (Q, U, G, I, L, R, S, T, E, V_CST).",
        "eml_summary_zh": "把「一輪」分解成向量 T =（U 使用者回合、G 生成軌跡、I 模型呼叫、L 外部回饋迴圈、R 重試、S 選擇／驗證、P 物理執行軌跡）。最乾淨的 single pass 是 U=1、G=1、R=1、L=0、S=0，排除的只有「行動→新證據→重規劃」，不排除自回歸的序列計算。LOOP 分五類（工具、環境、驗證器、重試、候選／選擇），能力分四層（intrinsic、elicited、system、product），任何「一輪完成」的宣稱都要投影到事件向量 (Q, U, G, I, L, R, S, T, E, V_CST)。",
        "eml_label_zh": "回合分解與外部無迴圈智能",
        "eml_primary_domain": "Computation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "Turn Decomposition Framework T = (U, G, I, L, R, S, P).",
          "Externally Loopless Intelligence (ELI): during a task the system cannot re-aim on external world, tool output, an independent verifier, another candidate or a retry.",
          "Interaction compression C_U = internal interactions / U.",
          "Loop taxonomy L = (L_T, L_E, L_V, L_R, L_C)."
        ],
        "eml_assumptions": [
          "Autoregressive generation is internal sequential computation, not an external loop.",
          "Selection over candidates is itself a capability source."
        ],
        "eml_claims": [
          "UserTurn ≠ PhysicalTurn; UserTurn ≠ ModelInvocation; ModelInvocation ≠ GenerationTrajectory.",
          "NoExternalLoop ≠ NoSequentialComputation.",
          "Pass@k ≠ Pass@1; BestOfN ≠ MedianExperience.",
          "InteractionCompression ≠ ComputationCompression; SameFinalQuality ≠ SameIntelligenceEfficiency."
        ],
        "eml_formalization": [
          "Single-pass condition: U=1, G=1, R=1, L=0, S=0 → Q_SP.",
          "P(at least one success in k rollouts) = 1 − (1 − p)^k.",
          "Unified event vector 𝔈 = (Q, U, G, I, L, R, S, T, E, V_CST)."
        ],
        "eml_predictions": [
          "Systems reported as 'one turn' will separate by orders of magnitude on (L, R, S, V_CST, E) once the event vector is recorded."
        ],
        "eml_falsification_conditions": [
          "If operational accounting of U, G, I, L, R, S adds nothing to predicting cost or quality beyond the visible answer, the decomposition is idle."
        ],
        "eml_known_limitations": [
          "Twelve invariants and a vocabulary; no measurement of its own — the vector is filled by Papers 02–09."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0101/",
      "json": "/ai/theory/THY-2026-0101/index.json"
    },
    {
      "id": "THY-2026-0102",
      "kind": "theory",
      "label": "μI — the minimum intelligent semantic execution unit (candidate theory)",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0102/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "μI is a task-relative semantic state transition z_t → z_{t+1} that satisfies seven conditions: task relevance, state change, causal contribution (Q(Y | μ) > Q(Y | do(μ=0))), non-decomposability at the chosen resolution, composability, realization independence and physical realizability. It is neither token, FLOP, neuron activation, layer nor thought; its minimality is resolution-, task- and observer-relative. Four candidate families (belief update, relation construction, constraint resolution, information gain) unify as a task-relevant, causally useful, non-redundant Δz. Gross vs effective counts give the semantic efficiency η_μ, and a cross-level map μI → ρ_C → ρ_P → ρ_T leads down to physical cost.",
        "eml_summary_zh": "μI 是相對於任務的語意狀態轉換 z_t → z_{t+1}，須滿足七個條件：任務相關、狀態改變、因果貢獻（Q(Y | μ) > Q(Y | do(μ=0))）、在指定解析度下不可再拆、可組合、實現無關、物理可實現。它既不是 token、FLOP、neuron activation、layer，也不是「想法」；最小性相對於解析度、任務與觀察者。四個候選族（信念更新、關係建立、約束消解、資訊增益）統一為任務相關、因果有用、語意不冗餘的 Δz。gross 與 effective 計數給出語意效率 η_μ，跨層映射 μI → ρ_C → ρ_P → ρ_T 一路向下接到物理成本。",
        "eml_label_zh": "μI——最小智能語意執行單位（候選理論）",
        "eml_primary_domain": "Cognitive Science",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "μI: z_t → z_{t+1} with z = (beliefs, relations, constraints, goals, uncertainty, procedures).",
          "Seven conditions C_μ = (C_T, C_S, C_C, C_N, C_K, C_R, C_P).",
          "N_μ^gross, N_μ^eff, η_μ = N_μ^eff / N_μ^gross; semantic work vector N_μ = (N_B, N_R, N_C, N_I, N_G).",
          "μI^obs (observer-reconstructed) vs μI^int (true internal); measurement ladder L0 behavioral proxy → L1 structured trace → L2 causal internal probe → L3 physical-semantic alignment."
        ],
        "eml_assumptions": [
          "A task-relative semantic solution state can be abstracted at observer level without claiming the model literally stores those fields.",
          "Counterfactual removal of a transition is at least conceptually testable."
        ],
        "eml_claims": [
          "Token ≠ μI ≠ FLOP ≠ NeuronActivation ≠ Layer ≠ Thought.",
          "SemanticActivity ≠ EffectiveIntelligentWork; SameAnswer ≠ SameSemanticExecution.",
          "SemanticWork ≠ PhysicalWork (type safety: semantic work is never a Joule)."
        ],
        "eml_formalization": [
          "Causal contribution: Q(Y | μI) > Q(Y | do(μI = 0)) in expectation.",
          "Semantic yield Y_μ = Q / N_μ^eff; Q = C_phys · η_{P→μ} · η_{μ→Q} (conceptual, not a physical equation).",
          "Signed contribution w_i^Q = ΔQ_i; equivalence class [μI] = {ρ | same pre-state, post-state and downstream function}."
        ],
        "eml_predictions": [
          "If μI is a good intermediate layer, adding N_μ improves prediction of efficiency, failure paths, scaffold gain and cross-architecture comparison over physical cost → quality alone (F5)."
        ],
        "eml_falsification_conditions": [
          "If N_μ never adds explanatory or predictive power over direct physical-cost → quality models, μI should be revised or eliminated (Paper 10, Falsifiable Claim 5)."
        ],
        "eml_known_limitations": [
          "Explicitly a candidate ontology: no claim that a unique neural or model-internal 'intelligence atom' exists; μI^obs ≈ μI^int is unproven.",
          "Experiment C (operational identification) has not been run."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0102/",
      "json": "/ai/theory/THY-2026-0102/index.json"
    },
    {
      "id": "THY-2026-0103",
      "kind": "theory",
      "label": "Cross-level triangulation and measurement grades",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0103/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Neuroscience has no '1 thought = N spikes' conversion; what it has is a cross-level proxy method: behavior → latent cognitive model → neural code → cellular events → physical implementation, each layer with its own units. IPM borrows five principles (level separation, latent inference, population over atomism, encoding–decoding duality, causal perturbation), extends Marr's three levels to five (task, semantic, algorithmic, physical events, thermodynamic), and defines cross-level triangulation for AI as E = (output, semantic, internal trace, ablation, hardware) evidence with a graded confidence in μI from D (behavioral) to A+ (physical-semantic alignment).",
        "eml_summary_zh": "神經科學沒有「1 個想法 = N 個 spike」的換算，有的是跨層代理量測：行為 → 潛在認知模型 → 神經編碼 → 細胞事件 → 物理實現，每層各有單位。IPM 借用五個原則（層級分離、潛變量推斷、群體優先於原子、編碼—解碼對偶、因果擾動），把 Marr 三層擴成五層（任務、語意、演算法、物理事件、熱力學），並為 AI 定義跨層三角化 E =（輸出、語意、內部軌跡、消融、硬體）證據，μI 的可信度分級從 D（行為）到 A+（物理—語意對齊）。",
        "eml_label_zh": "跨層證據三角化與測量等級",
        "eml_primary_domain": "Cognitive Science",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "Five layers L4 task achievement, L3 semantic/cognitive operation, L2 algorithmic realization, L1 physical events, L0 thermodynamic realization; L4 ≠ L3 ≠ L2 ≠ L1 ≠ L0.",
          "Cross-level triangulation X_L = (E_B behavioral, E_C cognitive-model, E_N neural, E_P perturbational); for AI X_AI = (E_O, E_S, E_I, E_A, E_H).",
          "Measurement grades D behavioral / C structured semantic / B internal correlation / A causal internal / A+ physical-semantic alignment.",
          "Eight borrowing principles: level separation, proxy discipline, model-mediated inference, distributed realization, causal perturbation, scale declaration, trial separation, grounding downward."
        ],
        "eml_assumptions": [
          "Latent quantities are scientific when they make observable predictions, have competitors, are falsifiable and accept intervention (the diffusion-decision-model template)."
        ],
        "eml_claims": [
          "CognitiveUnit ≠ NeuralEvent ≠ InformationBit ≠ PhysicalOperation; bits/spike ≠ cognitive bits; Decodable ≠ CausallyUsed.",
          "OutputRate ≠ InternalComputationRate (the ~10 bits/s behavioral throughput is not the brain's computation rate; 1000 output tokens are not 1000 intelligent events).",
          "EnsemblePerformance ≠ SingleEpisodePerformance; Proxy ≠ Ontology; BiologicalNeuron ≠ ANNNeuron."
        ],
        "eml_formalization": [
          "Conf(μI) = F(E_O, E_S, E_I, E_A, E_H) ∈ [0, 1]; report N̂_μ ± uncertainty with Grade_μ, never a falsely precise count.",
          "ValueOfComputation = ExpectedImprovement − Cost (resource-rational template for μI → Cost → Value)."
        ],
        "eml_predictions": [
          "Confidence that μI^obs ≈ μI^int rises only when behavioral, semantic, internal, causal and physical evidence converge."
        ],
        "eml_falsification_conditions": [
          "A μI account that survives behavioral evidence but is contradicted by ablation or hardware evidence must lose its grade, not be kept by verbal interpretation."
        ],
        "eml_known_limitations": [
          "Methodological borrowing only; no biological unit is equated with an AI unit and no measurement was performed."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0103/",
      "json": "/ai/theory/THY-2026-0103/index.json"
    },
    {
      "id": "THY-2026-0104",
      "kind": "theory",
      "label": "Energy accounting hierarchy and thermodynamic type safety",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0104/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Neural energetics builds energy bottom-up (membrane dynamics → ion flux → pump work → ATP → dissipation) and finds that a spike has no fixed energy and that most cortical signaling energy is spent on synaptic integration and state maintenance, not the visible pulse. IPM copies the discipline, not the numbers: energy is typed as E = (gross, baseline, marginal, attributed, thermodynamic minimum), a boundary and baseline rule must be declared, information per Joule is not intelligence per Joule, Landauer's kT ln 2 bounds erasure and is not the price of a μI, and Shannon or variational 'energies' never become physical Joules without an explicit mapping.",
        "eml_summary_zh": "神經能量學由下往上算能量（膜動力學 → 離子流 → 幫浦功 → ATP → 耗散），並發現一個 spike 沒有固定能量、皮質的訊號能量大半花在突觸整合與狀態維持而非顯眼的脈衝。IPM 複製的是紀律不是數字：能量分型為 E =（gross、baseline、marginal、attributed、熱力學下限），必須宣告邊界與基線規則，每焦耳資訊不等於每焦耳智能，Landauer 的 kT ln 2 只約束抹除、不是一個 μI 的價格，Shannon 或變分「能量」沒有明確映射前永遠不是物理焦耳。",
        "eml_label_zh": "能量帳本層級與熱力學型別安全",
        "eml_primary_domain": "Computation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "E_gross = ∫ P_system dt; E_base = ∫ P_baseline dt; E_marg = E_gross − E_base; E_attrib = E_marg + α·E_shared with a declared α.",
          "Energy boundary: accelerator / node / rack / data center / infrastructure / lifecycle; E = E(Boundary, BaselineRule, AttributionRule).",
          "E-grades: D estimated / C device telemetry / B node meter / A infrastructure meter / A+ marginal causal energy.",
          "Landauer distance D_L = E_actual / E_Landauer — an implementation distance, not an intelligence score."
        ],
        "eml_assumptions": [
          "Evolution and engineering optimize a Pareto set (energy, speed, reliability, robustness, adaptability), so minimum energy is not maximum utility."
        ],
        "eml_claims": [
          "Spike ≠ FixedEnergyUnit; Token ≠ FixedEnergyUnit; μI ≠ FixedEnergyUnit; SignalShape ≠ EnergyCost.",
          "GrossEnergy ≠ MarginalEnergy ≠ AttributedEnergy; EnergyComparison ⇒ SameBoundary.",
          "InformationPerJoule ≠ IntelligencePerJoule; LandauerBound ≠ ActualComputationCost; 1 μI ≠ kT ln 2.",
          "ShannonEntropy ≠ ThermodynamicEntropy and VariationalFreeEnergy ≠ PhysicalEnergy without an explicit mapping."
        ],
        "eml_formalization": [
          "Three efficiencies η_I/E = I/E, η_μ/E = N_μ^eff / E_marg, η_Q/E = Q / E_marg — never equated.",
          "Energy of a μI is a realization distribution P(E | μI, architecture, hardware, context, boundary)."
        ],
        "eml_predictions": [
          "Reports that give a single 'Joules per answer' without type and boundary will not be comparable across systems."
        ],
        "eml_falsification_conditions": [
          "If marginal, attributed and gross energies of the same task turn out to be interchangeable in practice, the typing is unnecessary."
        ],
        "eml_known_limitations": [
          "The pilot measured device energy only (E-Grade C); no marginal or attributed energy has been measured."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0104/",
      "json": "/ai/theory/THY-2026-0104/index.json"
    },
    {
      "id": "THY-2026-0105",
      "kind": "theory",
      "label": "Physical computation cost vector and computational spacetime",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E1",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0105/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Physical cost is the vector C_P = (typed operations, memory traffic by hierarchy, I/O, interconnect, memory residency, device occupancy, wall time, energy), not FLOPs. Computational spacetime is first a measure V_CST = ∫ R(t) dt = (V_C, V_M, V_N, V_S) over a resource field; the components may not be added before a declared normalization, and equal volume (8 GPU × 10 s = 1 GPU × 80 s) is not equal topology Θ_CST = (T_wall, T_serial, P_parallel, D_peak, M_peak, B_peak, Γ_comm). Roofline, memory-wall and data-movement results explain why same-FLOPs workloads differ in time and energy; a peak hardware footprint is a capacity barrier; CST grades run from D (spec estimate) to A+ (causal resource attribution).",
        "eml_summary_zh": "物理成本是向量 C_P =（分型的運算、按層級的記憶體流量、I/O、互連、記憶體駐留、裝置占用、wall time、能量），不是 FLOPs。計算時空首先是資源場的測度 V_CST = ∫ R(t) dt =（V_C、V_M、V_N、V_S）；未宣告正規化前各分量不可相加，體積相等（8 GPU × 10 s = 1 GPU × 80 s）不等於拓撲 Θ_CST =（T_wall、T_serial、P_parallel、D_peak、M_peak、B_peak、Γ_comm）相等。Roofline、memory wall 與資料搬移的結果解釋了為何相同 FLOPs 的工作在時間與能量上不同；峰值硬體占用是容量門檻；CST 等級從 D（規格推估）到 A+（因果資源歸因）。",
        "eml_label_zh": "物理計算成本向量與計算時空",
        "eml_primary_domain": "Computation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "C_P = (O, B_M, B_I, B_N, M_R, D, T, E); O typed by precision; nominal vs executed vs useful operations.",
          "V_M = ∫ M_resident dt (byte·s), V_C = ∫ D(t) dt (device·s); MemoryTraffic ≠ MemoryResidency.",
          "Normalized scalar V*_CST(Reference, Weights, Boundary); topology Θ_CST; peak footprint H_peak.",
          "Measurement confidence bundle G_M = (Grade_μ, Grade_E, Grade_CST); P_compute = (C_P, V_CST, Θ_CST, H_peak, E, Boundary_P, G_M)."
        ],
        "eml_assumptions": [
          "Attainable performance is bounded by min(P_peak, BW · arithmetic intensity) (Roofline)."
        ],
        "eml_claims": [
          "FLOPs ≠ PhysicalComputationalCost; SameFLOPs ≠ SameLatency ≠ SameEnergy; SameDeviceTime ≠ SameEnergy.",
          "SameCSTVolume ≠ SameCSTTopology; TotalResource ≠ PeakCapacityRequirement; MoreDevices ⇏ LowerLatency.",
          "ScalarCST ⇒ DeclaredNormalization; CSTComparison ⇒ SameBoundaryOrExplicitConversion; LowUtilization ≠ BadSystem."
        ],
        "eml_formalization": [
          "V_CST = ∫ R(t) dt with R = (r_C, r_M, r_N, r_S); V*_CST = ∫ Σ_j w_j r_j(t)/C_ref,j dt.",
          "Vector efficiency η_Q/CST = (Q/V_C, Q/V_M, Q/V_N, Q/V_S); Pareto dominance A ≻_P B."
        ],
        "eml_predictions": [
          "Controlling FLOPs will leave large independent variation in T, E, B_M, B_N and V_M across memory patterns and topologies (Falsifiable Claim 2)."
        ],
        "eml_falsification_conditions": [
          "If, FLOPs held fixed, time, energy, memory traffic and residency do not vary substantially across workloads, a single FLOPs cost model suffices."
        ],
        "eml_known_limitations": [
          "The pilot recorded wall time, device energy, peak VRAM, memory residency and utilization integrals for one GPU (CST-B); interconnect and memory traffic were not measured."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0105/",
      "json": "/ai/theory/THY-2026-0105/index.json"
    },
    {
      "id": "THY-2026-0106",
      "kind": "theory",
      "label": "Structured quality, hard gates and the specification–verification separation",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E1",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0106/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Quality is Q(Y | X, S, W, B_Q): output relative to task, specification, environment and boundary. It is measured as a structured vector (correctness, alignment, completeness, consistency, robustness, verifiability, provenance) in three layers — formal objective, structured semi-objective, human residual — with fatal conditions as a hard gate that soft quality cannot compensate. Coverage splits into specification, test and state coverage; mutation score measures test strength; verification and specification are separate axes (a perfect proof of the wrong theorem solves nothing); evaluator agreement is not truth; cost is not quality unless the specification makes it one. Quality evidence grades run from E (surface validity) to A+ (formal verification plus goal alignment).",
        "eml_summary_zh": "品質是 Q(Y | X, S, W, B_Q)：輸出相對於任務、規格、環境與邊界。以結構化向量（正確性、對齊、完整、一致、穩健、可驗證、來源）量測，分三層——形式化客觀、結構化半客觀、人類殘餘——致命條件是 hard gate，軟品質不能補償。覆蓋率拆成規格、測試與狀態覆蓋；mutation score 量測試強度；驗證與規格是兩個軸（完美證明了錯的定理什麼都沒解決）；評審一致不等於真；成本不是品質，除非規格把它變成品質。品質證據等級從 E（表面有效）到 A+（形式驗證加目標對齊）。",
        "eml_label_zh": "結構化品質、hard gate 與規格—驗證分離",
        "eml_primary_domain": "Evaluation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "Q_S = (Q_C, Q_A, Q_K, Q_R, Q_B, Q_V, Q_P), typed per domain; layers Q_L = (Q_F, Q_S, Q_H).",
          "Hard gate G_H(Y) = ∧ h_i(Y); coverage C_Q = (C_spec, C_test, C_state); mutation score MS = killed / non-equivalent mutants.",
          "Q_verified = Q_verification ⊗ Q_specification; math vector Q_math = (well-formedness, derivation validity, goal alignment, scope fidelity, axiom transparency, counterexample resistance).",
          "Quality object 𝔔 = (Q_S, G_H, C_Q, E_Q, Grade_Q, Conf_Q, U_Q, Boundary_Q); scalar Q* = Π_Q(𝔔 | task, projection rule)."
        ],
        "eml_assumptions": [
          "Evaluation oracles (tests, proof checkers, judge models, humans) are themselves fallible: ObservedQuality = F(TrueQuality, EvaluatorPower, Coverage)."
        ],
        "eml_claims": [
          "Quality ≠ IntrinsicScalar; SyntacticValidity ≠ SemanticCorrectness; CompileSuccess ≠ CorrectProgram; AllTestsPassed ≠ UniversalCorrectness.",
          "FormalVerification ≠ RealWorldGoalCorrectness; ProofValidity ≠ GoalEquivalence; ProofGrade ≠ GoalAlignmentGrade.",
          "FatalConstraintFailure ≁ SoftQualityTradeoff; PeakEpisode ≠ ReliableQuality; Length ≠ Completeness; Cost ≠ Quality.",
          "ObjectifiableFirst, HumanResidualLast."
        ],
        "eml_formalization": [
          "Robustness sensitivity S_R = ΔQ / d(x, x′); requirement coverage C_R = |satisfied| / n with typed importance."
        ],
        "eml_predictions": [
          "Instruments that fold output-format compliance into 'quality' will misreport task competence as failure on tasks the model actually solves."
        ],
        "eml_falsification_conditions": [
          "If a single universal quality scalar predicts downstream task success across domains as well as the structured object does, the structure is redundant."
        ],
        "eml_known_limitations": [
          "The first pilot showed the prediction in practice — two of three tasks had their 'quality' decided by JSON/Markdown obedience — but that is one instrument on one model, and the theory itself has not been tested beyond it."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0106/",
      "json": "/ai/theory/THY-2026-0106/index.json"
    },
    {
      "id": "THY-2026-0107",
      "kind": "theory",
      "label": "Binary residual quality measurement (IBQF / BRQM)",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0107/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "A 0–10 rating asks the respondent to perceive, build a reference, calibrate a scale, integrate dimensions and map to a number; the burden is highest exactly when the measured state is heaviest. BRQM instead takes many local, concrete, single-construct, non-numeric binary or pairwise answers b_i ∈ {0,1} and lets the measurement system reconstruct a latent multidimensional quality θ̂_H with Bradley–Terry / Thurstone / IRT-type models, adaptive item selection by information gain per human cost, blind and counterbalanced designs, and an explicit rater-disagreement structure — because binary observation is not binary phenomenon and disagreement is not error.",
        "eml_summary_zh": "0–10 評分要回答者同時感知、建參照、校尺度、整合維度、映射成數字；被測狀態最重時負擔正好最高。BRQM 改成收集大量局部、具體、單一構念、非數值的二元或成對回答 b_i ∈ {0,1}，讓測量系統用 Bradley–Terry／Thurstone／IRT 類模型重建潛在多維品質 θ̂_H，依「每單位人類成本的資訊增益」自適應選題，盲測與平衡設計，並明確保留評審分歧結構——因為二元觀測不是二元現象，分歧不是誤差。",
        "eml_label_zh": "二元殘餘品質測量（IBQF／BRQM）",
        "eml_primary_domain": "Cognitive Science",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "BRQM: 𝔔_H → {0,1}^N → θ̂_H; primitives b^abs ∈ {0,1} and b^pair ∈ {A, B}; skip = missing metadata, not a third value.",
          "Good-item conditions C_B = (local, single construct, concrete, temporally bounded, non-numeric).",
          "Adaptive selection i* = argmax E[IG_i] / C_H(i); stop when U_H < ε.",
          "Human residual object 𝔔_H^IBQF = (θ̂_H, Σ_H, N_obs, D_R, C_H, B_H, U_H, Grade_H); H-grades E uncontrolled rating … A+ cross-context validated."
        ],
        "eml_assumptions": [
          "A latent continuous quality exists behind local judgments (IBQF/FDCS micro-binary → macro-continuous emergence)."
        ],
        "eml_claims": [
          "BinaryObservation ≠ BinaryPhenomenon; HumanObservation ≠ HumanScaleConstruction; NumericRating = State + ScaleUse + Context.",
          "MeasurementBurden ≠ MeasuredQuality; Disagreement ≠ Error; MeanPreference ≠ PreferenceStructure; Reliability ≠ Objectivity.",
          "HumanResidual ⇏ HumanOverridesFormalTruth — the hard gate is applied first."
        ],
        "eml_formalization": [
          "P(A ≻ B) = σ(q_A − q_B) (Bradley–Terry); P(b_rij = 1) = σ(a_i θ_j − d_i + β_r), multidimensional λ_i^T θ_j, context-conditioned θ_j(c).",
          "C_rating = C_perceive + C_reference + C_scale + C_integrate + C_map; C_binary = C_local perceive + C_choose; C_binary, C_pair < C_rating is the hypothesis."
        ],
        "eml_predictions": [
          "With well-designed items, binary/pairwise adaptive protocols beat direct numeric rating on response time, consistency, dropout, predictive validity or fatigue in at least some settings (Falsifiable Claim 3)."
        ],
        "eml_falsification_conditions": [
          "If binary/pairwise protocols are worse than direct 0–10 rating on all of response time, consistency, dropout and predictive validity, the low-burden hypothesis must be revised."
        ],
        "eml_known_limitations": [
          "Not a clinical scale proposal; builds on EveMissLab's internal IBQF/MTF and FDCS theory (2025); Experiment B has not been run."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0107/",
      "json": "/ai/theory/THY-2026-0107/index.json"
    },
    {
      "id": "THY-2026-0108",
      "kind": "theory",
      "label": "Typed, versioned quality ontology for high-ambiguity artifacts",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0108/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Before any item is written, quality must be a typed space Q[domain, task, context, audience] = Q_core ⊕ Q_domain ⊕ Q_task, with a construct graph of dependencies and conflicts and a measurement itemization pipeline Task → Construct → Indicator → Item → Observation → Latent estimate. Any metric (BLEU, CLIPScore, aesthetic model score) is one projection of the space; a construct validity gate (coverage, discriminant validity, convergent evidence, context stability) guards against measuring the wrong thing precisely; the ontology is open — new constructs may be added from residual errors — but every revision is a version, and multimodal quality is not the mean of modality scores.",
        "eml_summary_zh": "在寫任何題目之前，品質必須先是有型別的空間 Q[領域、任務、情境、受眾] = Q_core ⊕ Q_domain ⊕ Q_task，附帶構念之間依賴與衝突的構念圖，以及測量條目化管線 任務 → 構念 → 指標 → 題目 → 觀測 → 潛在估計。任何指標（BLEU、CLIPScore、美學模型分數）都只是空間的一個投影；構念效度閘（覆蓋、區辨效度、收斂證據、情境穩定）防止「精準地量錯東西」；本體是開放的——可從殘餘錯誤新增構念——但每次修訂都是一個版本，而多模態品質不是各模態分數的平均。",
        "eml_label_zh": "高歧義成果的有型別、有版本品質本體",
        "eml_primary_domain": "Evaluation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "Core Q_core = (fidelity, coherence, completeness, robustness, usefulness, verifiability); domain schemas for text, image, music, story, design; multimodal coupling dimensions.",
          "Quality construct graph G_Q = (V_Q, E_Q); construct validity gate G_C; open ontology Q_{t+1} = Q_t ∪ {new construct} under versioning.",
          "High-ambiguity quality object 𝔔_HA = (Q_schema, G_Q, Q_F, Q_S, θ̂_H, Σ_H, D_R, B_Q, Version_Q)."
        ],
        "eml_assumptions": [
          "Subjective ≠ unstructured: large populations reliably detect structural failures even in creative artifacts."
        ],
        "eml_claims": [
          "HighAmbiguity ≠ Unmeasurable; Construct ≠ Indicator ≠ Item ≠ Metric; Reliability ≠ Validity.",
          "Fluency ≠ Factuality; Coherence ≠ Correctness; TechnicalQuality ≠ AestheticQuality; PromptSimilarity ≠ ImageQuality; Novelty ≠ Creativity; AestheticAppeal ≠ Usability.",
          "MultimodalQuality ≠ Mean(ModalityScores); CrossDomainComparison ⇒ SharedConstructBasis; OntologyRevision ⇒ Versioning; PreciseMeasurement ≠ CorrectConstructSelection."
        ],
        "eml_formalization": [
          "Quality fiber view Q = ∪_x Q_x over x = (d, τ, c, a); cross-task projection Π_{x→y} only over shared constructs.",
          "Creativity as a region (novelty, appropriateness, value, surprise, coherence), not novelty × usefulness."
        ],
        "eml_predictions": [
          "Benchmarks whose quality ontology drifts without versioning will produce longitudinal comparisons that are silently invalid."
        ],
        "eml_falsification_conditions": [
          "If a flat, unversioned checklist explains rater residuals and new failure modes as well as the typed construct graph, the ontology machinery is unnecessary."
        ],
        "eml_known_limitations": [
          "Schema proposals only; no domain ontology has been validated across populations."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0108/",
      "json": "/ai/theory/THY-2026-0108/index.json"
    },
    {
      "id": "THY-2026-0109",
      "kind": "theory",
      "label": "Scaffolding capability record: SSR, SDR, SCM and the ablation ladder",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "EXPERIMENTAL",
        "eml_evidence_level": "E2",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0109/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "A system is (model M, scaffolding vector S) with S = (tool, retry, multi-sample, verifier, environment, memory, planner). Single-pass quality Q_SP = Q(M, 0) and full quality Q_F = Q(M, S_F) define the scaffolding survival ratio SSR = Q_SP / Q_F, the dependence ratio SDR = 1 − SSR, the scaffold cost multiplier SCM_j = C_j^F / C_j^SP and marginal yields along a controlled ladder A0 native single pass → A1 extended budget → A2 multi-sample → A3 verifier → A4 tool/environment → A5 full agentic. The record also carries scaffold interactions (Shapley-style attribution), discarded and hidden work, and a capability vector (single-pass, loop, tool, verification, environment intelligence). A loop is not cheating; high dependence is a structure, not a defect — what destroys comparability is hiding the capability source and its physical cost.",
        "eml_summary_zh": "系統 =（模型 M，鷹架向量 S），S =（工具、重試、多樣本、驗證器、環境、記憶、規劃器）。單次品質 Q_SP = Q(M, 0) 與完整品質 Q_F = Q(M, S_F) 定義鷹架存活率 SSR = Q_SP / Q_F、依賴率 SDR = 1 − SSR、鷹架成本倍率 SCM_j = C_j^F / C_j^SP，以及受控階梯 A0 原生單次 → A1 放大預算 → A2 多樣本 → A3 驗證器 → A4 工具／環境 → A5 完整 agentic 上的邊際產率。紀錄同時帶鷹架交互作用（Shapley 式歸因）、被丟棄與隱藏的工作，以及能力向量（單次、迴圈、工具、驗證、環境智能）。LOOP 不是作弊；高依賴是一種結構不是缺陷——摧毀可比較性的是把能力來源與物理成本藏起來。",
        "eml_label_zh": "鷹架能力紀錄：SSR、SDR、SCM 與消融階梯",
        "eml_primary_domain": "Evaluation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "Scaffolding vector S = (S_T, S_R, S_N, S_V, S_E, S_M, S_P); Q_SP = Q(M, 0), Q_F = Q(M, S_F).",
          "SSR = Q_SP / Q_F; SDR = 1 − SSR; SCM_j = C_j^F / C_j^SP; marginal scaffolding yield Y_S,k = (Q_{k+1} − Q_k) / (C_{k+1} − C_k); scaffolding physical overhead ΔP_S = P_F ⊖ P_SP (typed).",
          "Capability vector I = (I_SP, I_L, I_T, I_V, I_E); Scaffolding Capability Record 𝔖_C = (𝔔_SP, 𝔔_F, ΔQ_S, SSR, P_SP, P_F, ΔP_S, G_S, φ, P_waste, Grade_S).",
          "S-grades E unknown harness / D declared components / C execution trace / B controlled ablations / A factorial interaction / A+ physical-causal attribution."
        ],
        "eml_assumptions": [
          "Conditions start from the same initial information (A0–A3); A4/A5 information gain is marked separately.",
          "Scaffold contributions interact (verifier needs candidates), so one-at-a-time ablation is not unique attribution."
        ],
        "eml_claims": [
          "Loop ≠ Cheating; SystemCapability ≠ ModelNativeCapability; Pass@k ≠ Pass@1; BestOfN ≠ SingleTrajectory.",
          "GenerationAbility ≠ VerificationAbility; ToolAccess ≠ ToolUtilizationIntelligence; ExternalMemory ≠ InternalLearning.",
          "ScaffoldingDependence ≠ SystemDefect; Loopless ≠ SuperiorByDefinition; InvisibleOutput ≠ ZeroCost; HiddenRetry ≠ FreeRetry.",
          "TransparencyOfCompute ≠ DisclosureOfPrivateReasoning; ScaffoldMeasurement ⇒ QualityAndCostTogether."
        ],
        "eml_formalization": [
          "Shapley-like attribution φ_i = Σ_A w(A)[Q(A ∪ {i}) − Q(A)] with sampling approximation and uncertainty U_φ.",
          "Brute-force region: ΔQ ≪ 1 with ΔC ≫ 1; selection waste ratio SWR = C_discarded / C_total; P_waste = Σ discarded attempts."
        ],
        "eml_predictions": [
          "A stable scaffolding response curve exists for a fixed model and task set; SSR < 1 on non-trivial tasks (Falsifiable Claim 4)."
        ],
        "eml_falsification_conditions": [
          "If Q_F ≈ Q_SP across most tasks, models and compute budgets, the native/system separation loses empirical necessity (F4)."
        ],
        "eml_known_limitations": [
          "First real data point (2026-09-03): SSR = 1.0 on three easy tasks with a 3.1× energy multiplier — a null result on an instrument whose quality axis was itself confounded; no factorial or Shapley analysis yet."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0109/",
      "json": "/ai/theory/THY-2026-0109/index.json"
    },
    {
      "id": "THY-2026-0110",
      "kind": "theory",
      "label": "The canonical intelligence event, Pareto comparison and no premature scalarization",
      "created_at": "2026-09-02",
      "updated_at": "2026-09-02",
      "values": {
        "eml_status": "PRELIMINARY",
        "eml_evidence_level": "E0",
        "eml_object_version": "0.1",
        "eml_canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0110/",
        "eml_provenance": {
          "source": "EveMissLab research collection: Intelligence Physical Metrology (真本體論13)",
          "extracted_by": "Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports",
          "extracted_at": "2026-09-11",
          "generator": "tools/extract_all.py",
          "claim_boundary": "status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports"
        },
        "eml_summary": "Intelligence is not a score of a model but an event 𝔍_IPM = (task object, quality object, semantic work object, physical computation object, scaffolding capability record, measurement metadata); the research object is the relation physical computation → effective semantic work → quality. Given a declared projection Q*, an intelligence yield vector (Q*/E_marg, Q*/V_C, Q*/V_M, Q*/B_M, Q*/B_N, Q*/T_wall) and semantic yields split efficiency into physical→semantic and semantic→outcome stages. Systems are compared on Pareto frontiers — single-pass, scaffolded, and their gap — under the rule 'vector before score, structure before average, uncertainty before false precision'; four capability archetypes (native, efficiently scaffoldable, compute-amplified, environment-coupled) are descriptive, not a ranking. A minimum reporting standard and a grade bundle (Q, μ, E, CST, S) make every claim carry its boundary and uncertainty. IPM is a metrology candidate, not a discovered natural constant.",
        "eml_summary_zh": "智能不是模型的分數，而是事件 𝔍_IPM =（任務物件、品質物件、語意工作物件、物理計算物件、鷹架能力紀錄、測量詮釋資料）；研究對象是關係 物理計算 → 有效語意工作 → 品質。在宣告的投影 Q* 下，智能產率向量（Q*/E_marg、Q*/V_C、Q*/V_M、Q*/B_M、Q*/B_N、Q*/T_wall）與語意產率把效率拆成「物理→語意」與「語意→成果」兩段。系統在 Pareto 前沿上比較——單次、鷹架化與其落差——遵守「能保留向量就不壓總分、能保留結構就不壓平均、能保留不確定性就不假裝精確」；四種能力原型（原生、可高效鷹架化、算力放大、環境耦合）是描述不是排名。最低報告標準與等級束（Q、μ、E、CST、S）讓每個宣稱帶著邊界與不確定性。IPM 是計量學候選框架，不是被發現的自然常數。",
        "eml_label_zh": "canonical intelligence event、Pareto 比較與不過早純量化",
        "eml_primary_domain": "Evaluation",
        "eml_program_id": "PRG-2026-0101",
        "eml_data_basis": "THEORY",
        "eml_definitions": [
          "𝔍_IPM = (𝔗, 𝔔_IPM, N_μ, P_compute, 𝔖_C, 𝔐) with 𝔗 = (X, S, W, B_Q, B_P) and 𝔐 = (uncertainty, versions, hardware, software, clock, provenance).",
          "Intelligence yield vector Y_I and semantic yields Y_μ, Y_Q/μ; two-stage efficiency η_{P→μ}, η_{μ→Q}.",
          "Pareto dominance A ≻_IPM B: 𝔔_A ⪰ 𝔔_B and every relevant cost axis ≤ with one strict, same task, schema, boundary and grade.",
          "Grade bundle G_IPM = (G_Q, G_μ, G_E, G_CST, G_S); IPM Minimum Reporting Standard v0.1 (task, quality, execution, physical, hidden work, measurement metadata)."
        ],
        "eml_assumptions": [
          "Cross-substrate comparison (GPU LLM, neuromorphic, symbolic, biological) is legitimate only with a shared task, a shared quality construct and semantic-equivalence evidence."
        ],
        "eml_claims": [
          "Intelligence ≠ TokenCount ≠ FLOPs ≠ BenchmarkScore ≠ OneUserTurn; Quality ≠ UniversalScalar.",
          "SemanticWork ≠ PhysicalWork ≠ EnergyOnly; SameQuality ≠ SamePhysicalCost ≠ SameSemanticWork ≠ SameQuality.",
          "Scalarization ⇒ DeclaredPolicy; Comparison ⇒ SharedBoundary; Measurement ⇒ Uncertainty; OntologyRevision ⇒ Versioning.",
          "IPM = MetrologyCandidate, not a discovered natural constant."
        ],
        "eml_formalization": [
          "Y_I = (Q*/E_marg, Q*/V_C, Q*/V_M, Q*/B_M, Q*/B_N, Q*/T_wall); brute-force region B_F(ε) = {c : dQ/dC < ε}; three frontiers F_Q/P, F_μ/P, F_Q/μ.",
          "Canonical comparison protocol: freeze task and quality schema → single pass → scaffolded → SSR/SDR/ΔP → N_μ where feasible → frontier → projection only if a decision needs it → grades and uncertainty → raw traces."
        ],
        "eml_predictions": [
          "Five falsifiable claims: token hypothesis, FLOPs sufficiency, binary burden, scaffolding separation, semantic intermediate utility."
        ],
        "eml_falsification_conditions": [
          "The framework is refuted piecewise: each of F1–F5 has its own condition, and μI in particular must earn predictive or explanatory utility or be dropped."
        ],
        "eml_known_limitations": [
          "A synthesis paper with no external references and no measurement of its own; the reporting standard has been applied once, to a three-task pilot."
        ],
        "eml_authors": [
          "Neo.K (EveMissLab)"
        ],
        "eml_ai_collaborators": [
          "Aletheia (GPT-5.6 Sol, OpenAI ChatGPT)"
        ]
      },
      "canonical_url": "https://evemisslab.com/ai/theory/THY-2026-0110/",
      "json": "/ai/theory/THY-2026-0110/index.json"
    }
  ],
  "snapshot": {
    "snapshot_id": "AI-SNAPSHOT-v0.1-fe85b9694a45",
    "created_at": "2026-09-11T05:00:27Z",
    "format_version": "0.1",
    "sedb_baseline": "v0.4B contract; static source content/ai/",
    "generator_version": "evemisslab-com ai_research 0.1",
    "object_count": 124,
    "relation_count": 499,
    "artifact_count": 58
  }
}