LFR / Physical–Digital Fidelity Auditing

Audit the distance between physical reality
and digital claims.

We audit whether sensors, records, algorithms and code preserve the physical meaning they claim to represent, in the chains we hold traces for: public grid records, operational telemetry, and our own code. We examine whether the physical object, the time basis and the evidential meaning survive as real-world processes are converted into data, transformed by software, and turned into claims or actions.

For asset review · technical due diligence · portfolio monitoring · underwriting evidence review

Buyer questions before renewal or underwriting See what we audit
Recomputable figures · Stated limits · No asset verdict Grid ledger recomputed daily · hash-anchored · Neutral

A correct calculation can still describe the wrong physical thing.

Software can be right about the arithmetic and wrong about the world. These are the seams where a digital claim quietly stops being about the object it names — and they are what we audit.

Object

A sensor can measure the wrong object.

Time

A timestamp can align the wrong event.

Identity

A market identifier can be mistaken for a physical asset.

Reproducibility

A reproducible pipeline can faithfully reproduce an invalid interpretation.

Validation

A validation gate can exist without validating the property it names.

We do not ask only whether the number is correct. We ask whether it is still about the right thing.

Audit, here. Audit here means testing whether a digital evidence chain is still about the physical thing it claims to represent. Statutory audit issues an opinion on financial statements; LFR asks an earlier question: is the evidence still about the right thing? We certify nothing and issue no assurance opinion.

Current field: grid-connected battery storage

Battery storage is our first full-scale physical–digital audit environment, because public grid records, market identities, dispatch traces, settlement data, on-site control signals and asset-level claims all coexist there at different observational layers. We make those layers explicit before they are allowed to collapse into a single story.

The GB Grid line Buyer questions The battery layer

From traces to a reviewable file

A dashboard reports one view. A data room describes the asset. LFR examines how multiple records hold together through time — and returns it in forms a review can actually file.

Basket view

Portfolio Status Sheet

Which records in a real asset basket require reconciliation.

 Single asset

Duty Trajectory Profile

How one asset’s visible operating rhythm has moved through time.

 Next question

Targeted Evidence Request Pack

What owner-held records should be requested to close an open question.
Travels with every review

Audit and recompute layer

How every released observation can be traced and checked by another competent party.

We do not turn an unresolved observation into a verdict. We turn it into a bounded question and a specific evidence request.

What you receive

Every engagement returns the same artifact class, judged on the same five properties.

  • A stated data boundary. Transfer method, retention, confidentiality and output scope are agreed before any dataset is received.
  • A versioned observation contract. What will be measured, on which channels, at what cadence, all fixed before the work begins.
  • Recomputable reported values. Every figure carries its null, its detection floor, and a recompute path a second party can walk without us.
  • Explicit limitations and unresolved contexts. What stayed below the measured floor, and what is marked unobserved because the asset never entered that region.
  • No health score and no asset verdict. We sell the measurement, never the ruling.

Current commercial line: GB Grid Evidence

Public grid records show part of what grid-connected assets are seen doing: their visible cadence, direction, volume, declarations, metered presence and operating context. LFR preserves those records continuously and turns them into authorised portfolio and asset reviews. Not asset health, dispatch causality, or a reliability forecast.

GB Grid / Public Physical LedgerEvidence pages

What can the public grid lenses actually support?

Dated system cards, anonymous public-duty archetypes, source capability, maturity horizons, reproducible methods, and the explicit boundary between a public question and an internal answer.

Observed evidence · Source correction · Claim boundary · Targeted evidence request
GB Grid / Balancing Cadence ExposureCadence lens
Anonymised fleet field from public GB balancing data: each dot one unit, no names; a marked boundary band shows where a public footprint can justify an evidence request.

Is one of yours quietly becoming a concentrated bearer?

Public BMRS accepted actions reveal how each asset is called: persistently, briefly, direction-switching, or episodically. The public lens shows the anonymised fleet field and what a footprint can justify asking; a named cadence profile is produced under authorisation, checked against your own instrumentation.

Cadence lens · Anonymised fleet field · Evidence requests · Named profile under authorisation
GB Grid / NESO Power-Flow FieldResearch demo

Does the support field still have more than one way to continue?

Public half-hour NESO data replayed as a moving power-flow organisation projection: the system-level field behind which individual assets carry their share.

Public specimens (2026-07-11 editions): Storage Fleet Duty Mirror · Grid Continuity Status Card    Paid surfaces: Portfolio Status Sheet · Targeted Evidence Request Pack

Current form: fileable HTML/PDF review · pilot-stage delivery · every figure recomputable.

View the public grid evidence How an authorised review works

What we preserve, and what we refuse

An audit is only as good as what it refuses to let through. These are the properties we carry across the chain, and the moves we will not make on the way.

What we preserve
  • Physical object identity
  • Time and event identity
  • Source provenance
  • Observation limits
  • Transformation and recomputation paths
  • Missing and unpublished evidence
  • Claim boundaries
  • Runtime evidence of what was actually consumed
What we do not do
  • We do not infer physical truth from data availability alone.
  • We do not turn public traces into private asset-health claims.
  • We do not treat reproducibility as proof of correct physical interpretation.
  • We do not allow language to exceed the evidence chain.

Real process → sensors and public records → digital representation → algorithms and code → claim, decision, action

From data observatories to embodied systems

The same fidelity problem appears wherever machines sense, model and act on the physical world. Today we audit public and operational data chains. The same architecture is built to extend to local sensors, physical runtime state, and eventually to embodied systems where actions must also be reconciled with what the world returns.

The first commercial implementation is GB Grid and storage. Other applications enter the product layer only when their channels, buyer and evidence workflow are real.

Method before interpretation

Six steps, identical in every domain. The GB Grid implementation documents each one against its exact public sources.

  1. Preserve the original record.
  2. Register what each channel can and cannot observe.
  3. Freeze definitions before reading the result.
  4. Derive within each source’s boundary.
  5. Test corroboration, contradiction and maturity.
  6. Publish the observation and the non-claim together.

No joined observation is stronger than its weakest identity or time bridge. Read the full method →

Bring us a real review question.

A portfolio under review. A submitted number without a clear evidence boundary. Two records that do not agree. A system whose dashboard does not explain what it has been carrying.

Discuss the review

We do not provide asset ratings, failure predictions, underwriting conclusions or generic dashboard development.

Observation Notes

Short records from work already underway. Interpretation follows observation; unresolved structure remains unresolved.

Observation 001
In a real BMS record, a whole-pack response body does not remain one fixed organisation through use.

BMS Research Observatory / a real-vehicle record / per-cell voltage and temperature channels

Observation 002
Scalar-matched battery pairs can occupy different observed response organisations.

Battery Organisation Observatory / Mirror Live Skin and Residual Atlas

Observation 003
Organisation changes around visible transition can remain observable without resolving a material mechanism.

ECI 72 Materials Workspace / DEM pipeline, 72 runs across six inclusion fractions

Observation 004
In a lab pack under WLTP, cells quietly lock into fixed roles, while flat capacity shows nothing.

Pack Reachability / a controlled lab pack, role-flow and possibility corridor

Observation 005
In local Alethia runtime probes, captured audio and phase-lock are recorded alongside field familiarity, tension, coherence, and permission.

Alethia / real-time audio runtime / 120 second baseline and intervention probes recorded 2026-05-09

Research Layer

Three observation eyes behind the products: a controlled lab cell, a lab pack under WLTP, and a real vehicle. Each evidenced on real data, none replacing SOH.

Controlled Cell

Lab Cell Residual

Cycler-collected 18650 cells observed for recurrence, residual, and role separation. The reproducible metrology basis for “same SOH, different organisation.”

Open controlled experiments
Controlled Pack

Pack Reachability

A 36-cell lab pack under WLTP, observed as role reachability and a possibility corridor: which cells carry load, whether substitutes remain, and where roles lock — invisible to flat pack capacity.

Runtime Prototype

Alethia Physical Interface

A local real-time runtime with recorded audio probes, phase-lock, memory-field influence, ethical tension, and permission gating; additional physical modalities remain in development.

Battery cells Battery packs Real-vehicle BMS Second-life evidence requests Materials / DEM Embodied systems / Alethia