6
Entries in the submitted prediction register
1
Framework paper under journal review
7
Companion drafts written, none submitted
18
Future-work items, two of them gates
No published work in this field yet. It is listed because it is part of the laboratory's scope, and entries will appear here as they are released.
Projective Correlation Theory: A Pregeometric Framework for Emergent Spacetime
A foundational framework proposing that spacetime emerges from underlying correlation structures. Includes a six-entry register of falsifiable commitments, Planck 2018 ΛCDM MCMC analysis, and spectral dimension signatures. The full plain-language account — the four primitives, the reconstruction theorem, the predictions, and the list of things it refuses to claim — is here.
⬡ Submitted to Foundations of Physics, July 2026 · under review · read the full account ↗
The Open Programme: What Is Proved, What Was Withdrawn, What Is Blocked
The current state of the research programme, written the way the internal records are written. Three campaigns retired the framework's flagship prediction; a three-line theorem then showed the retirement was necessary. Includes the two open gates, the paper series, the observational readiness audit, and the limits at the same resolution as the results.
⬡ Programme record · Read the full write-up ↗
Why the Gaussian Wins: One Map, Three Questions, One Answer
Iterated smoothing of a correlation kernel is heat flow. Its universality class, its renormalisation-group fixed point and its dynamical attractor turn out to be literally the same object — and a positive-definiteness requirement, long treated as bookkeeping, eliminates an entire family of nonlinear dynamics without consulting any data.
⬡ Mathematical notes · Read the full write-up ↗
The Correlation Budget: An Exact Bound on Perturbing a Covariance Structure
How much extra correlation can be injected between two regions before a covariance structure stops being one? There is a closed-form answer that fits on half a page. The algebraic bound is proved under stated hypotheses; the pre-registered numerical diagnostics around it produced one intended pass, two failures and one rule that could not be evaluated at all.
⬡ Short note + executed capsule · Read the full write-up ↗
Freeze First: Pre-registration for Computational Research
What gets hashed before a run, why every campaign is executed twice, and what happens when a decision rule turns out to reference a tolerance that was never defined. An honest account of five campaigns that all returned unwelcome verdicts — plus a checklist any small group could adopt on a Monday.
⬡ Method note · Read the full write-up ↗
Null Signal: A 70-Test Randomness Audit of a National Lottery
A case study in method calibration. Sixty model versions across three games, evaluated on a domain with an exceptionally strong fairness prior and a precisely specified random null. Complete record of what worked, what failed, and how model-selection leakage manufactures an edge — plus the one internally validated association, found in player choices rather than in the draws. Every ticket analysed still has a negative expected return.
⬡ Case study · Read the full write-up ↗
Hatchling: A Small Model That Learns After Training
The laboratory's own byte-level language model, with constant-memory recurrence instead of a growing cache. 1.4601 bits/byte at 9.09M parameters; a headline compression claim retired after it lost at every scale; a promotion threshold that turned out to be 999 times too large; a deletion path verified against the weights rather than a ledger; and a backdoor installed in our own model to show that a norm cap is the wrong shape of defence.
⬡ Laboratory record · Read the full write-up ↗
Research Paper Slot
How to read this list
These are working items, not announcements. A draft is a complete venue-neutral note that has not been submitted anywhere and has not been refereed; a gate is a missing object that blocks everything downstream of it; blocked means the block has been located precisely and is stated. The full reasoning is in the programme record: The Open Programme ↗
FW-1 — The Matter Interface
The single most consequential open problem: specify what matter fields are in this setting, how they couple to the reconstructed metric, and how they act back on the correlation structure. Opening it unlocks a stress tensor, candidate field equations, a derivation of the one calibrated coupling, and the horizon-thermodynamics sector.
○ Gate — open, no schedule
FW-2 — An Intrinsic Dynamics
A candidate has been postulated and audited: the structure evolves by its own emergence semigroup. It supplies a variational structure, an entropy arrow and an exact renormalisation fixed point — but not a field-theoretic action, which a functional renormalisation-group treatment requires and which waits on FW-1.
◑ Draft written · gate half-open
P3 — Conditional Gravitational Sector
Reciprocal-deformation representation, the vacuum obstruction to a pointwise coupling ansatz, weak-field correspondence, and an update-budget identity. The missing matter–environment interface is stated inside the draft rather than papered over.
◑ Drafted · venue-neutral · unsubmitted
P4 / P5 — Continuum Limit and the Local Critical-Surface Criterion
A convergence theorem at fixed regularisation, published together with the executed negative adjudication that retired the flagship step; and a local principal-symbol criterion for critical surfaces, explicitly conditional on a supplied continuation. Global horizons and the Hawking sector are deferred behind FW-1.
◑ Drafted · venue-neutral · unsubmitted
P2 — The Cosmological Sector
Not drafted, and will not be until three objects are derived: a physical bridge from correlation length to cosmological scale, a proper-time conversion, and an absolute placement of the transition relative to the observational pivot. Eliminating one assumption honestly exposed a clock contradiction; that is now the sector's stated block.
○ Blocked — block located and stated
NP-1 — Spin-Dependent Ringdown Timing
Twelve primary events plus one ordered reserve, frozen from the public catalogue. Acquisition executed under a value-blind contract: 49 files, 30.9 GB, every byte checksum-verified, with the counter for scientific values read standing at zero. The analysis stage is a separate contract and has not been authorised.
◑ Data frozen · analysis not authorised
FW-8 — Analogue Crossover-Location Test
The pre-registered protocol was rewritten after the retirement: it no longer looks for a jump, it measures where a smooth crossover sits across dimensionalities. What is missing is a laboratory — no examined open dataset supplies the required return probability with an independently calibrated horizon scale.
○ Protocol frozen · no platform adopted
Kord-X.Mem.OSS: Ambient Memory Architecture for Embedded AI Agents
Open-source memory system for long-horizon AI agents in embedded/ambient environments. Independent of the physics programme. Code available at github.com/kortxresearch.
◑ In Preparation
Research and AI disclosure
This page describes theoretical and experimental work. A preprint is not peer-reviewed, and planned work is not a production-ready product. AI-generated assistant responses may contain errors; verify scientific claims against the linked source materials. Research and correction enquiries: contact@kort-x.com.
A theory earns its place by telling the world exactly how to prove it wrong. We publish the target and hand over the arrows.