State estimation
Work with noisy, changing signals and compare estimation methods on the same observations.
Try the Black Box →AN INTERACTIVE SYSTEM DEMONSTRATION
LEXIUM
SYSTEMSA local-first experimental computing platform for state estimation, evidence-aware reasoning and governed simulation — built to keep evidence, uncertainty and action visibly distinct.
DEMO PUBLIC SURFACEVERIFIED LOCAL / PUBLIC SEPARATIONCANDIDATE 100K LOCAL BLACK BOX RUNNER
Website demonstration · Separate from the local operational master · A diagram is not a service health check.
FREEDOM · TRUTH · COHERENCE · RESPONSIBILITY
Less translation between observation and useful state.
Lexium is an experimental local-first computing environment. Its research asks whether more direct representations of state, relationships and change can reduce unnecessary processing while keeping evidence, uncertainty and permission boundaries visible.
Evidence-aware reasoning
Structure observations, relationships, competing explanations and preserved unknowns without pretending input evidence has been independently verified.
Understand Lexic →Bounded experimentation
Keep simulation, preparation and proposed actions separate from authority. A model result does not silently become an action.
See the architecture →CURRENT RESEARCH QUESTION
Representation-efficiency tests support the general shorter-path hypothesis. They do not yet establish that Lexic semantics causes all Lexium speed or estimator accuracy. That causal claim remains UNKNOWN.
How the parts fit together
Lexium organises reasoning, preparation and evidence around explicit governance. Select a component to inspect its role and limits.
GOVERNING PRINCIPLESALEX / GOVERNANCEreview · permission · trace
LEXICReasoning component / inspect evidenceVEILPermission and security boundaryATHENICPreparation and shadow / PREPARESTATE-SPACEControlled estimation experimentsNEXUSIntegration design / source contextEVIDENCEClaims, methods and limitationsConceptual relationships only. Athenic remains PREPARE; this map grants no execution authority. Inspect the component directory →
Choose a practical workspace
Test state estimation
Choose a repeatable dataset, inspect truth and noisy observations, then measure estimation error.
USE WHEN You want a reproducible state-estimation comparison rather than a marketing claim.
YOU GET Downloadable CSV datasets, MAE evidence and local black-box Lexium comparison without exposing the estimator source.
PUBLIC DATA · LOCAL PROPRIETARY RUNTIME
Run the MAE →Object evidence
Capture an image and inspect what is actually visible before identification or value is inferred.
USE WHEN You have an object or photo and need an evidence-first starting point.
YOU GET A retained image record; recognition and valuation still require a verified connector.
ANALYSIS CONNECTOR: NOT CONFIGURED
Open Prospector →Compare conflicting information
Supply observations that point to different answers and inspect how support accumulates.
USE WHEN Sources, notices or observations disagree and you need to see what supports each candidate.
YOU GET Ranked support, alternatives, limitations and next checks — not a guaranteed prediction.
LIVE LOCAL ENGINE · PUBLIC A$4 / READ NOT YET ENABLED
Open All-Seeing Eye →Quantify authorised system behaviour
Demonstrate local-first analysis of permitted telemetry and observable outputs without installing into or controlling the target system.
USE WHEN You want to understand the research direction for vehicles, machines or low-connectivity field environments.
YOU GET A static capability demonstration, example report and clear validation boundaries. There is no public runner or adapter access.
DEMONSTRATED CAPABILITY · NO PUBLIC EXECUTION
View OmniKey demo →Know what a result establishes
OBSERVATION
An input or measurement, with its source.
EVIDENCE
A record supporting a specific, bounded claim.
SIMULATION
A model output under stated assumptions.
PREPARED ACTION
A proposal awaiting review and permission.
AUTHORISED ACTION
An explicitly permitted action; verify its result.
UNKNOWN
A question the available evidence cannot resolve.
PASS, FAIL and UNKNOWN retain their meaning within the stated test. One synthetic run does not establish universal superiority. Inspect methods, evidence and limitations →
Know what is not finished.
The public demonstration is intentionally narrower than the local development system. These boundaries are part of the product, not footnotes.
Local Runner
The Black Box can execute Lexium locally here, but the Windows installer/signing path for outside researchers is not complete.
Medical / Bio
Bio twins, organ models and surgical simulation are research interfaces. They are not clinically validated and have no diagnostic authority.
Public paid execution
Public application rental and unfinished payment/fulfilment paths remain disabled until their delivery boundaries are verified.
Lexic causation
Compact representation has shown lower transport overhead, but Lexic semantics has not been proven to cause all estimator speed or accuracy results.
2.276× payload ratio
The recorded 100k request used 1,821,174 bytes for JSON and 800,056 bytes for LXS1. JSON was 2.276× the LXS1 wire payload — about 56.1% fewer bytes for LXS1. Scoped to this test; not a universal density claim.
Public / local separation
The public-safe build profile blocks local Command Centre, Dialogue, Analytics and operational APIs while preserving the local master.
Choose the evidence path that matches your question.
Try the Black BoxValidate a CSV locally, compare methods and export results.Inspect ProofSee scoped methods, results, hashes and limitations before accepting a claim.Understand the ArchitectureSee how reasoning, preparation, evidence and authority remain separated.CollaborateResearch, robotics, automotive and enterprise enquiries without emailing a sensitive dataset.