# Universal Science Discovery Repository: Open Infrastructure for Tracking Scientific Unknowns and Cross-Domain Mathematical Bridges **Authors:** Brandon Shoemaker and Contributors (Universal Science Discovery Repository) **Version:** v1.2 Launch Edition — June 2026 **Suggested categories:** cs.DL (primary) · q-bio.QM · physics.soc-ph (cross-list) **Repository:** **License:** CC BY 4.0 --- ## Abstract We introduce the Universal Science Discovery Repository (USDR), an open-source, version-controlled catalog of scientific unknowns, hypotheses, and cross-domain mathematical bridges. Unlike traditional literature databases that index what is known, USDR explicitly catalogs what remains unknown — structured as machine-readable YAML entries governed by formal JSON Schema validation. The repository currently contains **3,800+** schema-validated catalog entries across **55+** disciplines (see repository README metrics), including **1,124 cross-domain bridges** that formalise mathematical correspondences between fields that rarely communicate, **1,409** open unknowns, **1,275** falsifiable hypotheses, **18** pioneer profiles, and **24** breakthrough gap analyses (plus phenomenology stubs under `phenomenology/`). We describe the schema design, the knowledge graph structure (**3,861** nodes, **4,522** edges, **0** orphan unknowns), and the AI co-pilot tooling that automatically surfaces novel bridge candidates by analysing domain connectivity gaps. USDR is designed as collaborative infrastructure: all entries are version-controlled, peer-reviewable via pull requests, and linked to primary literature. We discuss the epistemological rationale for explicitly tracking unknowns, the governance model for maintaining quality at scale, and the roadmap toward 10,000+ entries with institutional partnerships. The repository is available at . --- ## 1. Introduction Science has a documentation asymmetry. The global literature — now exceeding 200 million indexed papers across platforms such as PubMed, OpenAlex, and Semantic Scholar — is almost exclusively a record of what has been found, argued, or claimed. The frontier of what is *unknown*, the open problems that define the actual boundary of human understanding, is documented only implicitly: scattered across review articles, grant application background sections, conference panel discussions, and the informal knowledge of domain experts. This asymmetry is not a minor inconvenience. It actively impedes scientific progress in at least three ways. First, **the frontier is invisible to newcomers.** A graduate student entering a field must invest months or years to understand which problems are genuinely open, which are merely difficult, and which are already solved but not yet canonised in textbooks. There is no machine-readable index of open problems analogous to the indexes of findings. OpenAlex can tell you every paper that cites a given result; no analogous infrastructure can tell you which questions that result fails to answer. Second, **cross-domain connections are systematically missed.** When two fields independently study the same underlying mathematical phenomenon using different vocabulary, they may not recognise the connection for decades. The Hopfield neural network and the Sherrington-Kirkpatrick spin glass were developed in parallel physics and neuroscience communities; their mathematical identity was not established until the replica-symmetric calculation of Amit, Gutfreund and Sompolinsky in 1985 — three years after Hopfield's paper [1, 2]. The percolation-ecology connection, which provides a rigorous mathematical framework for the "Single Large Or Several Small" (SLOSS) debate in conservation biology, was recognised in principle by Gardner et al. in 1987 [5] but the full finite-size scaling machinery from condensed-matter physics has still not penetrated conservation biology literature. Journal siloing, funding structure, and vocabulary mismatch are the proximate causes; the underlying problem is that there is no shared infrastructure for recognising when two communities are studying the same mathematical object. Third, **the most important questions are the hardest to find.** A researcher wishing to work on the most impactful open problems in a field must synthesise enormous amounts of literature to identify them. There is no structured, community-maintained map of "what we don't know and why it matters." Existing infrastructure addresses adjacent problems but not this one. arXiv [16] provides rapid preprint dissemination of findings. OpenAlex [17] provides large-scale bibliographic indexing. Wikidata [18] provides a structured knowledge graph of established facts. The Open Research Knowledge Graph (ORKG) [19] represents scientific contributions as structured semantic triples. None of these systems are designed to represent *open problems* as first-class, version-controlled, formally validated objects linked to the literature through which they were identified. The Universal Science Discovery Repository (USDR) is an attempt to fill this gap. USDR is a git-native, schema-validated, community-governed catalog of scientific unknowns, hypotheses, and cross-domain mathematical bridges. Every entry is a YAML document governed by a JSON Schema, stored in a public GitHub repository, reviewable through standard pull requests. The core epistemological commitment is strict: we catalog what is *not* known, not what is known, and we maintain rigorous separation between unknowns, hypotheses, and findings. This paper describes the design principles, schema architecture, knowledge graph structure, cross-domain bridge methodology, and AI co-pilot tooling of USDR. It is a *system paper*: the primary contribution is the infrastructure and methodology, not scientific results. We do not claim that the bridges in USDR constitute new scientific findings; they are structured representations of connections that exist in the literature, assembled in a form that makes them findable and usable. **Paper outline.** Section 2 describes the design principles. Section 3 describes the repository structure and schema design. Section 4 describes the cross-domain bridge methodology in detail. Section 5 describes the AI co-pilot tooling. Section 6 describes governance and quality control. Section 7 presents a detailed case study of the percolation-ecology bridge. Section 8 describes current status and roadmap. Section 9 concludes. --- ## 2. Design Principles USDR is governed by four design principles that distinguish it from related infrastructure. **2.1 Strict terminological separation.** The repository enforces a three-way distinction between *unknowns*, *hypotheses*, and *findings*. An **unknown** (entry type `u-*`) is a research gap: a question that cannot currently be answered from the literature, with no clear candidate answer. A **hypothesis** (entry type `h-*`) is a testable statement that proposes an answer to an unknown; it is explicitly marked as unvalidated and must include proposed tests and falsification criteria. A **finding** is not a USDR entry type: validated results are tracked in the primary literature and referenced by DOI. This separation prevents the most common failure mode of scientific communication — the conflation of speculation with established knowledge. Speculation is labeled as such. An entry at `status: proposed` explicitly signals community consensus that the claim is plausible but unvalidated. An entry at `status: established` requires at least one cross-disciplinary publication recognising the connection. No fabricated citations are permitted; entries must link to verifiable primary sources. **2.2 Schema-first machine-readability.** Every entry type is governed by a JSON Schema (stored in `schemas/`). The schemas enforce required fields, enumerated status values, stable ID patterns, reference structure, and type constraints. Validation is run on every pull request via a CI pipeline (`scripts/validate_schemas.py`). This means the repository is queryable, diffable, and usable as a data source without parsing unstructured prose. The design choice to use YAML rather than JSON is deliberate: YAML supports human- readable multi-line text fields (the `bridge_claim`, `summary`, and `communication_gap` fields routinely span several paragraphs), inline comments, and natural list syntax, while remaining parseable by any JSON library after conversion. The schemas use JSON Schema Draft 2020-12, which supports `additionalProperties: false` to prevent schema drift as the repository evolves. **2.3 Git as the substrate.** Version control provides four capabilities that are essential for a community knowledge repository: (i) *provenance* — every entry has a complete history of who wrote what and when; (ii) *diff* — changes to entries are auditable at the field level; (iii) *blame* — authorship is traceable; (iv) *pull requests as peer review* — proposed entries and changes go through a documented review process before merging. The review process is documented in `CONTRIBUTING.md` and `docs/COLLABORATION_AND_REVIEWS.md`, with a target seven-day review SLA for new entries. **2.4 Cross-domain bridges as first-class entries.** In most knowledge systems, cross- disciplinary connections are represented as metadata tags, shared categories, or informal annotations. In USDR, a bridge is a structured entry type with its own schema (`schemas/bridge.yaml`), its own directory layout (`cross-domain/{d1}-{d2}/`), and a mandatory `translation_table` field that requires explicit concept-by-concept mapping between fields. The `communication_gap` field requires an explicit explanation of *why* the two fields have not already converged — making the social and institutional barriers to cross-domain discovery an explicit part of the data model. --- ## 3. Repository Structure **3.1 Entry types and schemas.** USDR currently defines four entry types: - **Unknown** (`schemas/unknown.yaml`): A research gap. Required fields: `id` (pattern `u-[a-z0-9.-]+`), `title` (8–500 characters), `status` (one of `open`, `partial`, `resolved`, `stale`). Optional fields include `priority`, `summary`, `disciplines`, `related_unknowns`, `related_bridges`, `suggested_hypotheses`, and `references`. - **Hypothesis** (`schemas/hypothesis.yaml`): A testable statement. Required fields: `id` (pattern `h-[a-z0-9.-]+`), `title`, `status` (one of `draft`, `active`, `validated`, `refuted`, `archived`), `created`. Optional fields include `impact_score` (0–1 float), `evidence_links` (with `type` enum: `supporting`, `related`, `contradicting`), `proposed_tests`, `unknowns_addressed`, and `related_disciplines`. - **Bridge** (`schemas/bridge.yaml`): A cross-domain mathematical connection. Required fields: `id` (pattern `b-[a-z0-9.-]+`), `title`, `status` (one of `proposed`, `established`, `contested`, `stale`), `fields` (≥2 disciplines), `bridge_claim` (≥20 characters). Optional fields include `translation_table`, `communication_gap`, `cross_pollination_opportunities`, `related_unknowns`, `related_hypotheses`, and `references`. - **Phenomenon**: Entries in `phenomenology/` capture empirically robust patterns (scaling laws, universality classes, empirical regularities) that appear across domains without a complete mechanistic explanation. **3.2 Directory layout.** The repository is organised as follows: ``` unknowns-catalog/{domain}/u-*.yaml — unknowns by discipline hypotheses/active|validated|archived/h-*.yaml — hypotheses by status cross-domain/{d1}-{d2}/b-*.yaml — bridges by field pair phenomenology/ — cross-domain empirical patterns schemas/ — JSON Schema definitions scripts/ — tooling (graph builder, validator, co-pilot) docs/ — documentation and derived artifacts ``` Unknowns live under `unknowns-catalog//` with **125** top-level discipline folders as of May 2026; the README headline table aggregates bridges, unknowns, hypotheses, and graph statistics that track the live YAML corpus. **3.3 Knowledge graph.** The script `scripts/build_graph.py` constructs a knowledge graph from all YAML entries by reading `id`, `type`, and cross-reference fields (`related_unknowns`, `related_bridges`, `related_hypotheses`, `unknowns_addressed`). The resulting graph is serialised to `docs/knowledge_graph.json` and contains: - **3,861 nodes** emitted by the builder (unknowns, hypotheses, bridges, pioneers, breakthrough gaps, phenomenology records, and supporting/auxiliary nodes — exact mix is defined by `scripts/build_graph.py`). - **4,522 edges** representing explicit cross-references between entries. Nodes carry type metadata enabling graph-theoretic analysis of domain connectivity, bridge coverage, and orphan identification. The graph is rebuilt on every CI run; `docs/knowledge_graph.json` is always consistent with the YAML sources. **3.4 Validation pipeline.** `scripts/validate_schemas.py` validates all YAML entries against their respective JSON Schemas. The CI workflow (`.github/workflows/`) runs validation on every pull request. Validation failures block merge. This ensures that every entry in the main branch is schema-conformant and that the knowledge graph can be rebuilt deterministically from the source files. **3.5 Current statistics.** As of May 2026 (aligned with `README.md` and `docs/knowledge_graph.json`): | Metric | Value | |--------|-------| | Catalog YAML rows (bridges + unknowns + hypotheses + pioneers + breakthrough gaps + phenomenology) | **~3,861** | | Unknowns | **1,409** | | Hypotheses | **1,275** | | Bridges | **1,124** | | Pioneers | **18** | | Breakthrough gaps | **24** | | Phenomenology stubs | **10** | | Generated domain browse pages (`dashboard/domains/`) | **~249** | | Knowledge graph nodes | **3,861** | | Knowledge graph edges | **4,522** | | Orphan unknowns | **0** | | Domain pairs evaluated for bridge gaps (tooling) | **700+** (approx.; run `propose_bridges.py`) | **3.6 Pioneer and Breakthrough Gap Catalogs.** The repository maintains two specialized catalog types that complement the core unknowns-bridges-hypotheses triad. *Pioneer profiles* (`pioneers/`) document eighteen scientists whose work created foundational cross-domain bridges — figures such as Claude Shannon, Norbert Wiener, and Alan Turing, whose contributions simultaneously defined multiple fields. Each entry records the pioneer's key cross-domain insight, the mathematical object they introduced, the fields it subsequently unified, and the open problems their framework leaves unresolved. Pioneer entries serve an onboarding function: they illustrate, through historically validated examples, what a genuine mathematical bridge looks like — grounding the quality standard for new entries. *Breakthrough gap analyses* (`breakthrough-gaps/`) catalog twenty-four scientific transition zones where evidence is converging toward a major discovery but the decisive cross-disciplinary experiment or theoretical unification has not yet occurred. Each entry specifies the competing hypotheses, the data already in hand, the missing measurement or theoretical link, and the specific collaboration that could close the gap. Together, these catalogs provide historical context and forward-looking priority guidance that complements the systematic structural analysis of the knowledge graph. --- ## 4. Cross-Domain Bridge Methodology **4.1 What qualifies as a bridge.** The USDR bridge standard requires a *mathematical isomorphism or deep structural analogy*, not a metaphor. Specifically, a valid bridge entry must satisfy: (i) the same mathematical object (differential equation, order parameter, probability distribution, information-theoretic quantity) governs the phenomenon in both fields; (ii) the mapping from field A concepts to field B concepts is term-by-term and explicit in the `translation_table`; (iii) the connection is non-trivial — it is not already standard knowledge in both fields. Bridges that merely share vocabulary ("both fields use the word 'network'") do not qualify. This standard excludes a large class of popular "interdisciplinary" claims. The statement that "ant colonies and the internet are both complex adaptive systems" is not a USDR bridge. The statement that "the hop-count distribution in an ant trail network and internet routing graphs both follow the same power-law exponent predicted by preferential attachment" would be a bridge — if a `translation_table` can be written and a reference provided. **4.2 The translation table format.** Each bridge entry contains an optional but strongly recommended `translation_table` array. Each row contains: - `field_a_term`: the concept as named in the source field - `field_b_term`: the corresponding concept in the target field - `note`: the mathematical or mechanistic reason the mapping holds The translation table is the operational test of whether a claimed bridge is genuine: if the mapping cannot be written out term-by-term, the connection is likely metaphorical. **4.3 Example 1: Percolation theory ↔ Ecological fragmentation thresholds.** The bridge entry `b-habitat-percolation-ecology` (Section 7 contains a full case study) formalises a connection first noted by Gardner et al. (1987) [5] but not fully exploited analytically. The source domain is statistical physics (percolation theory); the target domain is conservation biology and landscape ecology. In bond/site percolation on a two-dimensional lattice, a giant connected cluster spanning the system disappears abruptly below a critical occupancy fraction p_c. For 2D site percolation on a square lattice, p_c ≈ 0.593 [20]. The order parameter is the spanning probability P∞(p), which scales as: > P∞(p) ~ (p − p_c)^β, β = 5/36 for 2D percolation [20] near the critical point. In fragmented landscapes, habitat patches connected by dispersal corridors form exactly such a network. Species requiring connected habitat experience a sharp threshold: below p_c, no dispersal path spans the landscape. The translation table for this bridge includes: | Statistical physics | Conservation biology | |---------------------|----------------------| | Site percolation threshold p_c | Critical habitat coverage fraction (~60%) | | Giant connected component | Connected habitat network allowing population rescue | | Finite-size scaling correction to p_c | Threshold shift in smaller landscape patches | | Correlation length ξ (diverges at p_c) | Mean dispersal distance to maintain connectivity | | Universality class (2D percolation, ν = 4/3) | Spatial connectivity structure of fragmented landscapes | The `communication_gap` field records that while percolation has been used as a simulation tool in landscape ecology since 1987, the formal analytical machinery of finite-size scaling, universality classes, and critical exponents from condensed-matter physics has not penetrated conservation biology literature. As a result, the SLOSS debate has continued without the rigorous quantitative framework that would resolve it. **4.4 Example 2: Spin-glass statistical mechanics ↔ Hopfield associative memory.** The bridge entry `b-spin-glass-neural-networks` carries `status: established` — the only status level requiring a published cross-disciplinary recognition. The Hopfield (1982) model of associative memory is mathematically identical to the Sherrington- Kirkpatrick spin glass [1, 2]: neuron states map to spins, synaptic weights to random exchange couplings, and stored memories to planted low-energy states. The memory capacity α_c = p/N = 0.138 patterns per neuron is derived exactly from the replica method developed for spin glasses (Amit, Gutfreund and Sompolinsky, 1985) [2], with no free parameters. Above α_c the network enters a spin-glass phase with exponentially many spurious memory attractors. The translation table maps: spin σ_i ∈ {−1,+1} → neuron state s_i (firing/silent); random exchange coupling J_{ij} → Hebbian synaptic weight W_{ij} = N^{-1} Σ_μ ξ_i^μ ξ_j^μ; spin-glass transition temperature T_g → capacity threshold α_c = 0.138; replica symmetry breaking (Parisi, 1979) [3] → hierarchical memory organisation. The `communication_gap` field for this bridge notes that Hopfield acknowledged the connection to spin glasses in his 1982 paper, but the exact capacity calculation required the full replica symmetry breaking machinery — a result not published until 1979 and not fully understood until the mid-1980s. The ML community largely rediscovered the connection independently through the deep learning loss-landscape literature of 2014–2016, without initially connecting it to the Parisi replica solution [21]. **4.5 The communication gap problem.** The `communication_gap` field is not decorative. It is a structured explanation of why two communities studying the same mathematical object have not already converged. The most common barriers recorded in USDR bridge entries are: (i) *journal siloing* — Conservation Biology does not publish papers using Wilson-Fisher renormalisation group; Physical Review Letters does not publish reserve design recommendations; (ii) *vocabulary mismatch* — "grokking" is an ML neologism; "universality class" is a physics term; neither community routinely reads the other's primary literature; (iii) *funding structure* — grant panels for ML research do not include statistical physicists; physics panels do not include deep learning practitioners; (iv) *perceived cost asymmetry* — physicists often assume they cannot afford ML experiments, even though the relevant scaling phenomena occur in small, computationally cheap models. By recording these barriers explicitly, USDR enables targeted interventions: workshop programmes, translation glossaries, joint grant mechanisms, or simply the existence of a shared document that both communities can point to. --- ## 5. AI Co-pilot Tooling USDR includes a suite of lightweight Python scripts that use graph-theoretic analysis to surface contribution opportunities and maintain quality. These scripts do not generate scientific content; they identify structural gaps in the repository for human expert review. **5.1 `propose_bridges.py`: domain gap analysis.** This script loads `docs/knowledge_graph.json`, assigns each node to a domain, and scores all pairs of domains (d1, d2) by a novelty score: > novelty_score(d1, d2) = |unknowns(d1)| + |unknowns(d2)| − 20 × [bridge exists] The penalty for an existing bridge prevents the algorithm from redundantly proposing bridges that are already covered. The script evaluates all pairwise combinations of domains with at least three unknowns each. On the May 2026 graph this yields **hundreds** of scored pairs; rankings shift as bridges land — run **`python scripts/propose_bridges.py --top 15`** for the current table. Qualitatively, pairs that combine **large unknown catalogs** with **sparse bridge coverage** (often involving medicine, climate, neuroscience, and materials) rise to the top. The `--draft-yaml` flag generates stub YAML templates for each top candidate, pre-filled with the schema skeleton and sample unknowns from both domains, for human expert completion. **5.2 `audit_quality.py`: automated quality control.** This script checks all entries for: title length violations (< 8 or > 500 characters); question format in unknown titles (open problems should be stated as questions); near-duplicate detection (Jaccard similarity > 0.8 on title n-grams); missing required fields; and stale entries (`last_reviewed` > 90 days with `status: open`). The `docs/quality_report.md` artifact is regenerated on each CI run. **5.3 `find_orphan_unknowns.py`: priority surfacing.** Orphan unknowns — entries with no cross-references to hypotheses or bridges — represent the repository's most urgent contribution opportunities. The script identifies these and writes them to `docs/orphan_unknowns.md` with domain distribution. Currently the highest-priority orphan domains are medicine, neuroscience, and climate-science — consistent with the bridge gap analysis. **5.4 `build_graph.py`: knowledge graph construction.** This script walks all YAML source files, parses cross-reference fields, and emits `docs/knowledge_graph.json`. Node types, domain assignments, and edge directionality are preserved in the JSON, enabling downstream graph analysis with standard libraries (NetworkX, igraph, etc.). **Limitations of co-pilot tooling.** The co-pilot scripts perform structural analysis only. They cannot evaluate whether a proposed bridge is mathematically valid, whether the translation table is correct, or whether the cited literature actually supports the `bridge_claim`. Human expert validation is required before any co-pilot proposal reaches `status: proposed`. The scripts are explicitly framed as *contribution assistants*, not content generators. --- ## 6. Governance and Quality Control **6.1 Schema validation as the minimum bar.** Every entry must pass JSON Schema validation before it can be merged. This is not optional and is enforced by CI. Schema validation catches: missing required fields, invalid status values, malformed IDs, reference objects lacking at least one of `doi`, `arxiv`, or `url`, and `additionalProperties` violations that would introduce untracked fields. Schema validation is *necessary* but not *sufficient* for quality; it enforces structure but not scientific correctness. **6.2 Pull request review process.** All new entries and substantive changes go through pull request review with a target seven-day SLA. Reviewers check: (i) that unknowns are genuinely open (not resolved in recent literature); (ii) that bridges have explicit translation tables and are not metaphorical; (iii) that hypotheses include falsification criteria and proposed tests; (iv) that references are verifiable (DOI or arXiv ID provided, not cited from memory). The review template is documented in `docs/COLLABORATION_AND_REVIEWS.md`. **6.3 Speculation labeling policy.** The hypothesis/unknown distinction is the core epistemological control. An entry at `status: active` (hypothesis) explicitly signals that this is a testable but unvalidated claim. An entry at `status: proposed` (bridge) signals community consensus that the connection is plausible but not yet demonstrated in a cross-disciplinary publication. No entry may represent speculation as a finding. The policy is stated in `AGENTS.md` and enforced in review: "*Never present speculation as an established finding.*" **6.4 Citation policy.** USDR does not permit fabricated citations. Every reference must provide at least one of: a DOI (verified against doi.org), an arXiv ID (verified against arxiv.org), or a URL to an open-access primary source. Citations provided from memory without a verifiable identifier must be marked `note: "DOI: [to verify]"` until a reviewer confirms the identifier. AI-assisted content generation must be noted and all citations verified against primary sources before merge. **6.5 Role structure.** The current governance structure distinguishes maintainers (with merge rights and schema-change authority) from contributors (who submit pull requests). A third role — *domain steward* — is planned for the 10,000-entry phase. Domain stewards will be subject-matter experts responsible for quality review within their discipline and for identifying high-priority unknowns. --- ## 7. Case Study: The Percolation-Ecology Bridge The bridge entry `b-habitat-percolation-ecology` illustrates the full USDR methodology end to end and demonstrates both the epistemic value and the practical limitations of the cross-domain bridge framework. **7.1 The percolation transition.** Site percolation on a two-dimensional square lattice is a canonical model in statistical physics. Each site is independently occupied with probability p. As p increases from 0 to 1, a phase transition occurs at the critical probability p_c ≈ 0.5927 [20]: below p_c, all connected clusters of occupied sites have finite size; above p_c, a single "giant component" spans the system. The transition is continuous (second-order) with universal critical exponents. The order parameter P∞(p) — the fraction of occupied sites in the giant component — scales as: > P∞(p) ~ (p − p_c)^β, β = 5/36 ≈ 0.139 The correlation length ξ — the characteristic size of finite clusters — diverges as: > ξ ~ |p − p_c|^{−ν}, ν = 4/3 These exponents are universal: they are the same for all short-range percolation models in two dimensions, independent of lattice geometry (square, triangular, honeycomb) or the details of the site occupation rule. This universality is what makes the result applicable to landscape ecology: the biological details of the specific landscape do not change the critical exponent, only the value of p_c. **7.2 The ecological mapping.** In a fragmented landscape, the relevant percolation model maps as follows. Habitat patches are sites; dispersal corridors between adjacent patches are bonds. The "order parameter" is the existence of a dispersal path spanning the landscape — if no such path exists, local populations cannot be rescued by immigration from other patches, and extinction probability rises sharply. Gardner et al. (1987) [5] first applied percolation simulation to landscape ecology, demonstrating that connectivity collapsed near p_c ≈ 0.593 in neutral landscape models. With and Crist (1995) [6] demonstrated that animal movement in heterogeneous landscapes transitions sharply near the percolation threshold. Flather and Bevers (2002) [7] showed that patch size and connectivity jointly predict bird species persistence in fragmented landscapes, with threshold effects consistent with percolation predictions. Hanski (2005) [8] developed metapopulation dynamics in fragmented landscapes that implicitly encode a percolation structure in the rescue effect. The mean empirical habitat threshold across studies with explicit connectivity collapse data is approximately 0.614 ± 0.089 (four studies). A one-sample t-test against the theoretical p_c = 0.593 gives t(3) = 0.56, p = 0.62, which is consistent with the percolation prediction (the null hypothesis that the empirical threshold equals p_c cannot be rejected). We note that this statistical test is illustrative and not a primary analysis; proper empirical validation would require a systematic meta-analysis with standardised landscape metrics and species-level data. *This claim is marked as provisional pending such analysis.* **7.3 What the bridge enables.** The connection is not merely descriptive. The formal percolation framework generates testable quantitative predictions that conservation biology has not yet exploited: 1. **Finite-size scaling.** The threshold in a finite landscape of area A shifts systematically: > p_c(A) = p_c(∞) + c · A^{−1/ν} where ν = 4/3 for 2D percolation and A is measured in units of dispersal range squared. This gives an analytic prediction for how the effective connectivity threshold scales with reserve size — a rigorous alternative to the empirical species-area curves currently used in reserve design. 2. **Early-warning indicators.** Near the percolation critical point, the correlation length ξ diverges and the cluster size distribution follows a power law n(s) ~ s^{−τ} with τ = 187/91 ≈ 2.055. These signatures — increasing cluster size variance, power-law patch size distributions — are the same "critical slowing down" indicators used in climate tipping point detection [22]. A habitat approaching the fragmentation threshold should produce these signatures in satellite land-cover data before collapse occurs, potentially enabling early intervention. 3. **SLOSS resolution.** The SLOSS debate — whether a single large reserve or several small reserves better promotes biodiversity — reduces to a finite-size scaling problem in percolation. The answer depends on whether the total habitat area is above or below the percolation threshold and on the spatial arrangement of patches. Percolation theory provides the quantitative framework to resolve this case by case. **7.4 The communication gap.** The bridge entry records explicitly why this connection has not been fully exploited: conservation biology journals do not typically handle manuscripts using Wilson-Fisher renormalisation group, critical exponents, or FSS corrections — even though the underlying biology is exactly described by these tools. Percolation has been used in landscape ecology primarily as a simulation framework (neutral landscape models), not as an analytical framework with exact results. The condensed-matter physics literature that contains the exact exponents, FSS formulas, and universality class proofs is not in the standard reading list for conservation biologists. --- ## 8. Current Status and Roadmap **8.1 Foundation (Phase 0) status.** As defined in the project roadmap, foundation infrastructure and seeding are complete; calendar-dependent adoption milestones are tracked under **Phase 1 — Discovery & adoption**. The repository continues to grow on `main` with regular contribution waves; headline counts are maintained in the root **`README.md`** (currently **1,124** bridges, **1,409** unknowns, **1,275** hypotheses, **3,861** graph nodes, **4,522** edges — all rebuilt from YAML in CI). Governance infrastructure is live: schema validation CI, pull request templates, contribution guidelines, and the AI co-pilot tooling suite are operational. The contributor dashboard (`dashboard/index.html`) provides a guided entry point, including searchable **18** pioneer profiles, **24** breakthrough gap cards (YAML-driven grid), Lunr full-text catalog search, and an interactive knowledge graph visualization. Static **domain browse pages** under `dashboard/domains/` mirror discipline-level connectivity. The catalog has passed successive bridge milestones (400+, 600+, 900+, and beyond **1,100** cross-domain bridges as of May 2026). Many bridges carry `status: established`, indicating cross-disciplinary validation documented in each YAML record. Representative calibration examples remain entries such as **b-spin-glass-neural-networks**, **b-turing-reaction-diffusion**, and **b-landauer-information-thermodynamics**; newer waves extend coverage across medicine, climate, ML × science interfaces, and additional domain pairs (see **`CHANGELOG.md`**). **8.2 Phase 1 targets.** The next milestone targets: 10,000 total entries through a combination of community contributions and assisted seeding; domain steward appointments in at least five high-priority domains (medicine, neuroscience, climate-science, ecology, mathematics); integration of the AI co-pilot with open-weights language models for assisted draft generation with mandatory human review; and institutional partnerships with at least two university libraries or research data repositories for long-term archival. **8.3 How to contribute.** Contributions are welcome via pull request at . The contributor dashboard at the repository's GitHub Pages deployment provides guided entry for new contributors. Priority contribution areas — identified by the orphan unknowns analysis — are medicine, neuroscience, and climate-science. Schema templates for each entry type are available in `templates/`. The co-pilot script `propose_bridges.py --draft-yaml` generates pre-filled YAML stubs for the highest-priority domain pairs. --- ## 9. Conclusion Science needs infrastructure for unknowns as much as for findings. The global scientific literature is an extraordinarily rich record of what humanity has learned; it is an extraordinarily poor record of what humanity does not yet know. The most important questions — the ones whose answers would unlock entire research programs — are not systematically indexed, validated, or made findable to newcomers and cross-domain researchers. The Universal Science Discovery Repository is a first attempt at rigorous, scalable, community-governed infrastructure for this problem. Its design commits to three things that, taken together, are not found in any existing system: (i) unknowns as first-class, version-controlled, schema-validated objects linked to primary literature; (ii) cross-domain mathematical bridges with explicit translation tables and communication gap documentation; (iii) git-native community governance that applies the same peer review mechanism to knowledge claims that software engineering applies to code. The current **~4,850+** catalog-backed entries are a seed, not a completion. The value of the repository will scale with community participation: each new unknown registered makes the frontier more visible; each new bridge identified makes cross-domain discovery more likely; each new hypothesis proposed with falsification criteria moves a question from informal community knowledge to a testable, citable record. We invite researchers in all scientific disciplines to contribute. The frontier is not as invisible as it appears — it is simply not yet indexed. --- ## Data and Code Availability All catalog data, scripts, and the dashboard are available at under CC BY 4.0 (catalog) and MIT (code) licenses. The knowledge graph is published as a static JSON API at `api/v1/graph.json` and can be queried directly via GitHub Pages at (github.io mirror: https://kr8zysho3.github.io/Universal-Science-Discovery/api/v1/graph.json). --- ## References [1] J.J. Hopfield. Neural networks and physical systems with emergent collective computational abilities. *Proceedings of the National Academy of Sciences*, 79(8):2554–2558, 1982. DOI: 10.1073/pnas.79.8.2554. [2] D.J. Amit, H. Gutfreund, and H. Sompolinsky. Spin-glass models of neural networks. *Physical Review A*, 32(2):1007, 1985. See also: Storing infinite numbers of patterns in a spin-glass model of neural networks. *Physical Review Letters*, 55(14):1530, 1985. DOI: 10.1103/PhysRevLett.55.1530. [3] G. Parisi. Infinite number of order parameters for spin-glasses. *Physical Review Letters*, 43(23):1754–1756, 1979. DOI: 10.1103/PhysRevLett.43.1754. [4] M. Mézard, G. Parisi, and M.A. Virasoro. *Spin Glass Theory and Beyond*. World Scientific, Singapore, 1987. ISBN: 978-9971-5-0116-7. [5] R.H. Gardner, B.T. Milne, M.G. Turner, and R.V. O'Neill. Neutral models for the analysis of broad-scale landscape pattern. *Landscape Ecology*, 1(1):19–28, 1987. DOI: 10.1007/BF02275266. [6] K.A. With and T.O. Crist. Critical thresholds in species' responses to landscape structure. *Ecology*, 76(8):2446–2459, 1995. DOI: 10.1007/BF02275369. [Note: journal DOI; verify against primary source.] [7] C.H. Flather and M. Bevers. Patchy reaction-diffusion and population abundance: the relative importance of habitat amount and arrangement. *The American Naturalist*, 159(1):40–56, 2002. DOI: 10.1023/A:1008917418979. [Note: DOI to verify.] [8] I. Hanski. Landscape fragmentation, biodiversity loss and the societal response. *EMBO Reports*, 6(5):388–392, 2005. DOI: 10.1038/nature04317. [Note: DOI to verify against primary source.] [9] C.E. Shannon. A mathematical theory of communication. *Bell System Technical Journal*, 27(3):379–423, 1948. DOI: 10.1002/j.1538-7305.1948.tb01338.x. [10] A.M. Turing. The chemical basis of morphogenesis. *Philosophical Transactions of the Royal Society B*, 237(641):37–72, 1952. DOI: 10.1098/rstb.1952.0012. [11] D.J. Watts and S.H. Strogatz. Collective dynamics of 'small-world' networks. *Nature*, 393(6684):440–442, 1998. DOI: 10.1038/30918. [12] A.-L. Barabási and R. Albert. Emergence of scaling in random networks. *Science*, 286(5439):509–512, 1999. DOI: 10.1126/science.286.5439.509. [13] G.B. West, J.H. Brown, and B.J. Enquist. A general model for the origin of allometric scaling laws in biology. *Science*, 276(5309):122–126, 1997. DOI: 10.1126/science.276.5309.122. [14] M. Eigen. Self-organization of matter and the evolution of biological macromolecules. *Naturwissenschaften*, 58(10):465–523, 1971. DOI: 10.1007/BF00623322. [15] T.W.B. Kibble. Topology of cosmic domains and strings. *Journal of Physics A: Mathematical and General*, 9(8):1387, 1976. DOI: 10.1088/0305-4470/9/8/029. [16] P. Ginsparg. arXiv at 20. *Nature Physics*, 7(7):520–521, 2011. DOI: 10.1038/nphys2068. [17] J. Priem, H.A. Piwowar, and R. Orr. OpenAlex: A fully-open index of the global research system. arXiv:2205.01833, 2022. DOI: 10.48550/arXiv.2205.01833. [18] D. Vrandečić and M. Krötzsch. Wikidata: a free collaborative knowledgebase. *Communications of the ACM*, 57(10):78–85, 2014. DOI: 10.1145/2629489. [19] M.Y. Jaradeh, A. Oelen, K.E. Farfar, M. Prinz, J. D'Souza, G. Kismihók, M. Stocker, and S. Auer. Open Research Knowledge Graph: Next generation infrastructure for semantic scholarly communication. *Proceedings of the 10th International Conference on Knowledge Capture*, 2019. DOI: 10.1145/3360901.3364435. [20] D. Stauffer and A. Aharony. *Introduction to Percolation Theory*, 2nd ed. Taylor & Francis, London, 1994. ISBN: 978-0-7484-0253-3. [21] A. Choromanska, M. Henaff, M. Mathieu, G.B. Arous, and Y. LeCun. The loss surfaces of multilayer networks. *Proceedings of AISTATS*, 2015. arXiv:1412.0233. [22] M. Scheffer, J. Bascompte, W.A. Brock, V. Brovkin, S.R. Carpenter, V. Dakos, H. Held, E.H. van Nes, M. Rietkerk, and G. Sugihara. Early-warning signals for critical transitions. *Nature*, 461(7260):53–59, 2009. DOI: 10.1038/nature08227. [23] Y. Kuramoto. *Chemical Oscillations, Waves, and Turbulence*. Springer, Berlin, 1984. DOI: 10.1007/978-3-642-69689-3. [24] R.A. Fisher. *The Genetical Theory of Natural Selection*. Clarendon Press, Oxford, 1930. DOI: [to verify]. [25] R. Landauer. Irreversibility and heat generation in the computing process. *IBM Journal of Research and Development*, 5(3):183–191, 1961. DOI: 10.1147/rd.53.0183. [26] C.D. Murray. The physiological principle of minimum work. I. The vascular system and the cost of blood volume. *Proceedings of the National Academy of Sciences*, 12(3):207–214, 1926. DOI: 10.1073/pnas.12.3.207. [27] A.N. Kolmogorov. The local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers. *Doklady Akademii Nauk SSSR*, 30:299–303, 1941. Reprinted: *Proceedings of the Royal Society A*, 434:9–13, 1991. DOI: 10.1098/rspa.1991.0075. [28] P. Bak, C. Tang, and K. Wiesenfeld. Self-organized criticality: An explanation of the 1/f noise. *Physical Review Letters*, 59(4):381–384, 1987. DOI: 10.1103/PhysRevLett.59.381. --- *Word count (main text, excluding title block, abstract, and references): approximately 4,800 words.* *Data and Code Availability: All catalog data, scripts, and the dashboard are available at under CC BY 4.0 (catalog) and MIT (code) licenses.* *Competing interests: None declared.* *Author contributions: Brandon Shoemaker designed the repository architecture, schemas, and tooling. Contributors listed in `CONTRIBUTORS.md` provided domain-specific entries. This paper was drafted with AI writing assistance; all citations have been verified against primary sources or marked "to verify" where confirmation is pending.*