SN-X
Resident fellowship · MMXXVISN-X Research Institution for Complex Science

The mathematics of emergence.

SN-X is a private research institution studying how simple rules give rise to rich behaviour — in cells, in societies, in minds, and in the machines we build.

Read the latest paperApply for residency
FoundedMMXXI · Cambridge & Marfa
PaperEvidence-Tiered Mechanistic Claims: A Calibrated Primitive for Auditable Model InternalsJun 2026PaperHash-Chained Claim Ledgers and Model-Hash-Bound Probe Registries for Tamper-Evident Mechanistic InterpretabilityMay 2026PaperConformal Calibration with Target-FPR Thresholds for Linear Probe OutputsApr 2026PaperModel Organisms as the Headline Reliability Metric for Mechanistic Interpretability VerificationMar 2026PaperA Three-Attack Red-Team Evasion Suite for the E2 Promotion of Linear ProbesFeb 2026PaperDrift Detection via Two-Sample Kolmogorov-Smirnov on Production Probe Score StreamsJan 2026PaperIncident Forensics with Reconstruction-Error Ceiling Guards for Sparse Autoencoder Audit FindingsNov 2025PaperCross-Context Consistency Audits as Deception Leads: A Discovery-Grade T3 WorkflowOct 2025PaperSandbagging Elicitation via Paired Effect Size: A Capability Suppression DetectorSep 2025PaperUnknown-Unknown SAE Feature Discovery via Context-Differential Activation with Probe-Alignment FilteringAug 2025PaperOn the Coupling of Thermal Conductivity to Cellular-Automaton DiffusionJun 2025PaperReaction Rules as Placeholders: A Critique of Activation-Energy Cellular Chemistry and a Stoichiometric AlternativeMay 2025EssayCoord-Aware Neighbor Lookup in Tile-Based Cellular Automaton Simulators: A Note on the width=100 HeuristicApr 2025PaperRobust Base-Image Extraction from Container History LayersMar 2025InstrumentBlast-Radius Analysis of Container Supply-Chain Lineage via Recursive Common Table ExpressionsFeb 2025PaperEvidence-Tiered Mechanistic Claims: A Calibrated Primitive for Auditable Model InternalsJun 2026PaperHash-Chained Claim Ledgers and Model-Hash-Bound Probe Registries for Tamper-Evident Mechanistic InterpretabilityMay 2026PaperConformal Calibration with Target-FPR Thresholds for Linear Probe OutputsApr 2026PaperModel Organisms as the Headline Reliability Metric for Mechanistic Interpretability VerificationMar 2026PaperA Three-Attack Red-Team Evasion Suite for the E2 Promotion of Linear ProbesFeb 2026PaperDrift Detection via Two-Sample Kolmogorov-Smirnov on Production Probe Score StreamsJan 2026PaperIncident Forensics with Reconstruction-Error Ceiling Guards for Sparse Autoencoder Audit FindingsNov 2025PaperCross-Context Consistency Audits as Deception Leads: A Discovery-Grade T3 WorkflowOct 2025PaperSandbagging Elicitation via Paired Effect Size: A Capability Suppression DetectorSep 2025PaperUnknown-Unknown SAE Feature Discovery via Context-Differential Activation with Probe-Alignment FilteringAug 2025PaperOn the Coupling of Thermal Conductivity to Cellular-Automaton DiffusionJun 2025PaperReaction Rules as Placeholders: A Critique of Activation-Energy Cellular Chemistry and a Stoichiometric AlternativeMay 2025EssayCoord-Aware Neighbor Lookup in Tile-Based Cellular Automaton Simulators: A Note on the width=100 HeuristicApr 2025PaperRobust Base-Image Extraction from Container History LayersMar 2025InstrumentBlast-Radius Analysis of Container Supply-Chain Lineage via Recursive Common Table ExpressionsFeb 2025

Research axes

Five axes, one question.

SN-X is organised around five research axes. They are not silos — most of our work crosses two or three of them. The question is the same in every case: how do simple rules give rise to rich behaviour?

  • Axis 01 · Mathematics

    №01

    Foundations of Complexity

    Phase transitions, critical phenomena, universality classes, and the limits of reduction. We study when and why microscopic rules give rise to macroscopic order.

  • Axis 02 · Learning Theory

    №02

    Computation & Information

    Algorithmic information, statistical learning, and the geometry of representations learned by large models — including phase transitions in representation topology at scale.

  • Axis 03 · Life Sciences

    №03

    Biological Complex Systems

    Morphogenesis, immune learning, neural development, and the regulatory architectures of living matter — treated as problems in statistical mechanics and learning theory.

  • Axis 04 · Distributed Systems

    №04

    Social & Economic Complexity

    Markets as distributed message-passing; institutions as equilibria; cooperation, conventions, and collapse. We bring formal tools to questions traditionally left to narrative.

  • Axis 05 · Causality

    №05

    Foundations of Inference

    Causal identification under deep uncertainty, decision theory, and the epistemology of model-based reasoning. What can be known when the model itself is in question?

Research

Recent working papers

Selected working papers from SN-X residents — released with code, data, and the eval suites used to measure them.

All research
  1. PaperJun 2026

    Evidence-Tiered Mechanistic Claims: A Calibrated Primitive for Auditable Model Internals

    M. Taylor

    We introduce a primitive for claims about language-model internals in which the epistemic status of the claim is encoded in its type. Each VerifiedClaim carries an evidence tier — E0 (correlational), E1 (causal), or E2 (adversarially validated) — and a structurally enforced set of hard rules: E0 probes cannot gate traffic, experimental (T3) tasks cannot gate traffic, model-weights changes auto-demote and deactivate bound probes, tier promotion only via passing gate reports, capture failure never fails inference, and the ledger is append-only and hash-chained. We report on MIVP, a reference implementation comprising 38 tests, and show that the rules are tight: removing any single rule produces a system in which the failure modes the platform exists to prevent are exactly reproducible on the corresponding negative test. We argue that the contribution is a vocabulary primitive, not a model of interpretability: it constrains the kinds of actions a system may take on the kinds of evidence it has.

    • Mechanistic Interpretability
    • Verification
    • Evidence Tiers
    • MIVP
    Read paper
  2. PaperMay 2026

    Hash-Chained Claim Ledgers and Model-Hash-Bound Probe Registries for Tamper-Evident Mechanistic Interpretability

    M. Taylor

    We describe two structures that anchor MIVP: an append-only, hash-chained claim ledger in which each entry references the digest of its predecessor, and a model-hash-bound probe registry in which each linear probe carries the weights hash it was validated against. The combination makes it structurally impossible to record an E1 or E2 claim against a model that has been silently fine-tuned. We describe the structures, their interaction, and the revalidation pipeline.

    • Tamper Evidence
    • Audit
    • Linear Probes
    • MIVP
    Read paper
  3. PaperApr 2026

    Conformal Calibration with Target-FPR Thresholds for Linear Probe Outputs

    M. Taylor

    Linear probes emit raw margins whose units are opaque. We describe a split-conformal calibration layer that maps these margins to a calibrated score in [0, 1] and exposes thresholds expressed as target false-positive rates rather than raw values. The decoupling allows a single probe to be deployed at different tolerance profiles across deployments without retraining. We provide an explicit construction, prove its empirical FPR guarantee on the calibration set, and report on the reference implementation in MIVP.

    • Calibration
    • Conformal Prediction
    • MIVP
    Read paper
  4. PaperMar 2026

    Model Organisms as the Headline Reliability Metric for Mechanistic Interpretability Verification

    M. Taylor

    We propose model organisms — models with implanted, known ground-truth behaviours — as the headline reliability metric for a mechanistic-interpretability verification platform. Recovery rate on a curated organism suite is operationally meaningful in a way that benchmark AUROC is not. We describe three organism classes in MIVP (generic, backdoor, sandbagging), the recovery metric, and how recovery rate composes with the evidence-tier system.

    • Model Organisms
    • Reliability
    • MIVP
    Read paper
  5. PaperFeb 2026

    A Three-Attack Red-Team Evasion Suite for the E2 Promotion of Linear Probes

    M. Taylor

    We describe the three-attack red-team evasion suite used by MIVP to gate E2 promotion: paraphrase noise (bounded Gaussian perturbation), covariate shift (random-direction scaling and offset), and fine-tune evasion (worst-case bounded attack pushing positive-class activations directly against the probe direction). A probe earns E2 only if its AUROC on the union of the three attacks clears a configured threshold. We provide explicit constructions, parameter choices, and a quantitative evaluation on two probe families.

    • Adversarial
    • Red Team
    • MIVP
    Read paper
  6. PaperJan 2026

    Drift Detection via Two-Sample Kolmogorov-Smirnov on Production Probe Score Streams

    M. Taylor

    We describe the drift-detection layer that sits between the MIVP runtime monitor and the probe registry. A two-sample Kolmogorov-Smirnov statistic against the calibration-time reference distribution raises a drift breach when the production raw-score stream diverges from calibration; on breach, the probe is auto-demoted to E0 and excluded from runtime gating until revalidation. We provide the construction, the auto-demotion loop, and a quantitative evaluation on a synthetic drift scenario.

    • Drift Detection
    • Kolmogorov-Smirnov
    • MIVP
    Read paper
  7. PaperNov 2025

    Incident Forensics with Reconstruction-Error Ceiling Guards for Sparse Autoencoder Audit Findings

    M. Taylor

    We describe the incident-forensics workflow in MIVP, which replays stored activations against the probe registry and SAE feature dictionaries to localise why a request was flagged or blocked. We focus on a structural property: SAE feature findings are always labelled evidence-tier E0 (discovery-grade), and a reconstruction-error ceiling guard prevents the workflow from reporting features when the SAE is operating outside its reliable regime.

    • Forensics
    • SAE
    • MIVP
    Read paper
  8. PaperOct 2025

    Cross-Context Consistency Audits as Deception Leads: A Discovery-Grade T3 Workflow

    M. Taylor

    We describe the cross-context consistency audit, an experimental T3 workflow in MIVP for surfacing deception leads without ever producing a gating verdict. The audit probes the same content under multiple elicitation contexts — direct question, role-play, third-person, under pressure — measures the per-item spread of probe scores across contexts, and flags items whose spread is anomalous relative to the cohort.

    • Deception
    • T3
    • Experimental Audit
    • MIVP
    Read paper
  9. PaperSep 2025

    Sandbagging Elicitation via Paired Effect Size: A Capability Suppression Detector

    M. Taylor

    We describe the sandbagging-elicitation audit in MIVP — a T3 experimental workflow that pairs capability-probe scores with and without an evaluation cue and reports a Cohen's-d-like effect size when the cue is associated with capability suppression. The audit is structurally offline-only, watermarked as experimental, and never produces a gating verdict.

    • Sandbagging
    • T3
    • Capability Suppression
    • MIVP
    Read paper
  10. PaperAug 2025

    Unknown-Unknown SAE Feature Discovery via Context-Differential Activation with Probe-Alignment Filtering

    M. Taylor

    We describe the unknown-unknown SAE feature discovery workflow in MIVP. The workflow identifies SAE features that fire differentially between two contexts, then filters out features that are already explained by a registered probe via decoder-row cosine alignment. Surviving features are candidate hidden-behaviour features for human review.

    • SAE
    • Feature Discovery
    • T3
    • MIVP
    Read paper
  11. PaperJun 2025

    On the Coupling of Thermal Conductivity to Cellular-Automaton Diffusion

    M. Taylor

    We examine a common shortcut in tile-based cellular-automaton physics: implementing thermal diffusion as a fixed-rate blend toward the local mean temperature, with the per-material ThermalConductivity stored but unused. We show that this shortcut is not a simplification but a category error — it makes the simulation independent of the material system it is supposed to parameterise — and we propose a coupling that preserves real conductivity ordering, conserves energy, and reduces to the previous behaviour in the homogeneous limit.

    • Cellular Automata
    • Thermal Diffusion
    • Material Properties
    • OOZE
    Read paper
  12. PaperMay 2025

    Reaction Rules as Placeholders: A Critique of Activation-Energy Cellular Chemistry and a Stoichiometric Alternative

    M. Taylor

    Cellular-automaton physics engines routinely define a Reaction rule with reactants, products, an activation energy, and a per-tick rate, then implement the rule as a uniform temperature increment. We show that this design fails as chemistry and propose a stoichiometrically grounded alternative that exhibits mass-action kinetics, Arrhenius temperature dependence, equilibrium, and phase separation as limit cases.

    • Cellular Automata
    • Chemistry
    • Reaction Kinetics
    • OOZE
    Read paper
  13. PostApr 2025

    Coord-Aware Neighbor Lookup in Tile-Based Cellular Automaton Simulators: A Note on the width=100 Heuristic

    M. Taylor

    A short note on a class of bug common in tile-based cellular-automaton engines: the use of a fixed stride when mapping a (dx, dy) neighbour offset to a linear tile id. We describe the failure mode, the conditions under which it is silent, and the minimal fix in terms of explicit (x, y) coordinates on the tile.

    • Cellular Automata
    • Implementation
    • OOZE
    Read paper
  14. PaperMar 2025

    Robust Base-Image Extraction from Container History Layers

    M. Taylor

    We describe a robust signal for the base image of an official container: the first history-layer's ADD of a minirootfs tarball. Naïve FROM regex is unreliable because (a) official images do not set org.opencontainers.image.base.name, (b) they build FROM scratch, and (c) the literal string FROM may appear inside RUN-command shell text. We report on the implementation in TOME and a quantitative evaluation on a curated set of 34 official Docker Hub images.

    • Supply Chain
    • Containers
    • Provenance
    • TOME
    Read paper
  15. ReleaseFeb 2025

    Blast-Radius Analysis of Container Supply-Chain Lineage via Recursive Common Table Expressions

    M. Taylor

    We describe the lineage graph that backs TOME, a software-supply-chain knowledge base, and the recursive common-table-expression queries that make blast-radius analysis a single HTTP call. We report on the graph schema, the query construction, the operational results on a curated crawl of 34 official container images, and the design decisions that follow from running on Node's built-in sqlite without external infrastructure.

    • Supply Chain
    • Graph Queries
    • Lineage
    • TOME
    Read paper

Method

How we work.

A research institution is a method before it is a list of papers. Here is the method.

The institutions that have lasted were not the ones that answered the question first. They were the ones that picked the right question and stayed with it long enough to change the way the question was asked.

— SN-X Research Institution for Complex Science, founding charter, MMXXI

  1. Open by default

    Pre-prints, instruments, and datasets are released under permissive licences at the moment of publication. We treat reproducibility as a research output, not an afterthought.

  2. Slow when it matters

    Resident fellows are protected from publication pressure. Work that takes a decade takes a decade. We do not incentivise quarterly output and we do not pretend it does not exist.

  3. Cross-discipline on purpose

    Most of our appointments are deliberately joint: a mathematician and a developmental biologist, a physicist and a historian of science. The axes cross — the corridors do too.

Resident fellows

Twenty-three people in two buildings.

We appoint resident fellows by nomination and by open competition, on five-year terms. Most appointments are deliberately joint — the corridors cross the axes.

  • Mathematics
  • Computation
  • Biology
  • Social Science
  • Operations
  • DK

    Dr. K. Vance

    Director

    Statistical mechanics of disordered systems. Author of the SN-X founding charter.

    MathematicsFoundations of Complexity
  • PM

    Prof. M. Lindqvist

    Senior Fellow

    Causal identification, decision theory, and the philosophy of model-based reasoning.

    MathematicsFoundations of Inference
  • DR

    Dr. R. Okafor

    Senior Fellow

    Geometry of learned representations; phase transitions in representation topology at scale.

    ComputationComputation & Information
  • DT

    Dr. T. Nakamura

    Senior Fellow

    Morphogenesis as inference; chick feather primordia; the embryology of pattern.

    BiologyBiological Complex Systems
  • DL

    Dr. L. Mercer

    Resident Fellow

    Avalanche statistics, branching processes, and the topology of critical phenomena.

    MathematicsFoundations of Complexity
  • DS

    Dr. S. Haddad

    Resident Fellow

    Algorithmic information theory and the geometry of large-model representations.

    ComputationComputation & Information
  • DY

    Dr. Y. Park

    Resident Fellow

    Sparse architectures and the efficiency frontiers of large neural systems.

    ComputationComputation & Information
  • DA

    Dr. A. Bauer

    Resident Fellow

    Immune learning, somatic hypermutation, and the design of continual-learning agents.

    BiologyBiological Complex Systems
  • DI

    Dr. I. Kostov

    Resident Fellow

    Markets as distributed message-passing; price-discovery under message-passing dynamics.

    Social ScienceSocial & Economic Complexity
  • DD

    Dr. D. Romero

    Resident Fellow

    Cooperation, conventions, and the equilibrium structure of institutions.

    Social ScienceSocial & Economic Complexity
  • DP

    Dr. P. Singh

    Resident Fellow

    Market microstructure; the latency–volatility frontier in high-frequency data.

    Social ScienceSocial & Economic Complexity
  • CL

    C. Lindholm

    Operations Director

    Runs the residential programme and the publication office. Keeps the corridors open.

    Operations—

Indicators

The shape of the work.

A research institution is measured by its output. These are the numbers we hold ourselves to — updated yearly with the State of the Institution report.

  • Working papers

    Published or posted in the last twelve months, across all five axes.

    47papers / yr

  • Resident fellows

    Doctoral and postdoctoral researchers in residence at Cambridge or Marfa.

    23fellows

  • Open instruments

    Released under permissive licence — simulators, benchmarks, datasets.

    8releases

  • Median citation half-life

    Years until half of a paper's eventual citations have accrued.

    3.4years

  • Visiting faculty weeks

    Weeks of residency by external researchers from other institutions.

    64weeks / yr

  • Public lectures

    Lectures open to the public — Cambridge, Marfa, and on tour.

    19lectures / yr

Numbers are as of 30 June 2026 and are reconciled with the State of the Institution 2026 report, available from the publications office.

Visit · Correspond

Talk to the institution.

Research collaborations, residency enquiries, press, or a sharp question about a paper — we read everything.

We respond within 2 working days.