A permanent evidence layer for agriculture.

I

Agriculture has data everywhere. Evidence nowhere.

Modern agriculture generates enormous quantities of data each season, and very little durable evidence. The same problem repeats in five recognisable patterns. They compound each other.

Agriculture has no shared memory layer for operational evidence.

II

Why biologicals first.

Proof is designed to support agricultural evidence broadly over time. It launches with biologicals and biostimulants in arable systems. The choice is deliberate.

Biological products represent one of the fastest-growing and most variable areas of agriculture. Outcomes are highly context-dependent. Treatments that perform well in one season, soil type or rotation routinely fail to repeat in another. The category currently faces several specific evidence problems:

• Inconsistent evidence quality across manufacturers and trial programmes
• Fragmented trial methodologies that do not compare across farms
• Varying protocols even within a single product’s published evidence
• Survivorship bias toward published positive outcomes
• Limited operational transparency about how trials were run
• Low comparability between datasets held by different parties

Farmers increasingly want evidence grounded in real operational conditions on working farms, rather than isolated marketing claims. Biologicals are the right entry point for trusted evidence infrastructure. High stakes, high variability, low existing transparency.

The long-term scope extends far beyond biologicals. Each later category opens only when the schema can carry it without distorting the records already in place.

III

Farm evidence should be permanent, structured and protected.

Proof turns real on-farm observations into structured, append-only records that can be inspected and cited over time. A locked record cannot be silently rewritten; corrections appear as visible addenda. A record stands for what it is: one outcome, in one field, in one season, under stated conditions.

The public projection preserves the evidence. The private envelope protects the farms and organisations behind it. The two surfaces are designed in tandem, and neither is an afterthought.

Public evidence preserved. Private farm data protected.

IV

Arable systems first. Agriculture in time.

Proof v1 launches in arable systems, with biological and biostimulant trials as the first record category. The schema, the trust model and the governance rules are built to extend across the breadth of agricultural work. Regenerative arable extends naturally as the density of arable records grows.

Later phases reach into livestock, dairy, horticulture, viticulture, agroforestry, mixed farming, controlled environment agriculture, perennial systems and emerging enterprises. Each enterprise is added only when the schema can carry it without distorting the records already in place.

Biologicals are the first wedge. Arable is the first scope. Agriculture is the long-term scope.

V

What a Proof Record is.

A Proof Record is one outcome from one field, structured under a published schema, locked at a point in time, and published as a public redacted projection. Every record has a unique identifier, a canonical hash, a six-state trust pip line, an evidence manifest and a permanent ledger.

A record is one outcome, not a verdict. A single record does not say that a product works in general; it says that under these specific conditions, this outcome was measured and structured according to rules anyone can inspect. Comparable records, taken together, form the basis on which broader claims can be cautiously made.

Evidence types.

A record’s evidence manifest can carry items across the categories below. Each item is linked to the section of the record it supports, carries a per-item hash, and is preserved with provenance metadata. Raw files remain in the private envelope; the public projection shows the manifest structure with metadata.

PhotographsApplication photos, batch photos, in-field observations. EXIF metadata preserved; raw imagery stays private.

Laboratory reportsSoil tests, tissue analyses, product specifications, residue and assay reports. PDFs and structured exports both accepted.

ProtocolsPre-defined trial design, application plan, measurement plan. Protocol-locked records pre-register the design before outcome.

Field observationsStructured notes, dated observations, in-season events that affected the trial.

MeasurementsYield files, weighbridge tickets, measurement sheets, instrument outputs. Tabular formats preferred.

Application recordsRates, timings, equipment, environmental conditions at application. Linked to treatment arm.

Supporting documentationProduct labels, manufacturer trial references, supplier correspondence, conflict declarations.

VI

How a record is made.

Five stages take a real on-farm trial from a draft to a permanent, publicly inspectable record. Each stage is a publishing rule, not an editorial judgement.

A reader can stop at any stage of a record’s history and see what was captured, how it was structured, when it was locked, what was published, and what changed afterwards. None of those answers depend on Proof’s say-so. Each is part of the record itself, carried as ledger metadata.

VII

What is public. What is private.

The public record carries the evidence. The private envelope carries the farm. The split is structural, defined in the schema, and identical for every record published on the platform.

Proof makes evidence inspectable without making farms exposed.

VIII

System architecture.

Proof is built as structured evidence infrastructure, not as a document repository. The architecture is organised around append-only evidence, immutable ledger events, controlled vocabularies, role-based access controls and per-record audit checkpoints.

1. Structured capture models. A published schema defines the shape of every record. Free-text capture is bounded by structured fields, controlled vocabularies and required evidence.
2. Locked relational records. A locked record is a permanent, append-only object. Corrections appear as visible addenda; the underlying record is not rewritten.
3. Immutable ledger events. Every lock, addendum, Evidence Query, conflict declaration and access event is written to a ledger as an immutable event. The ledger is the audit surface.
4. Public projections. Each locked record produces a public redacted projection. The projection is the canonical public surface for that record, with a stable URL and an OG card.
5. Private evidence envelopes. Sensitive material, including exact farm identity, raw files, GPS boundaries, invoices and contributor legal identity, sits in a private envelope behind access controls, separate from the public record.
6. Audit checkpoints. Periodic ledger checkpoints carry the cumulative hash of the platform’s record state, making silent edits detectable from outside the system.

The architecture is designed for preservation across years and decades, not quarters. Schema versions are tracked. Migration of older records into newer schema versions is itself logged. The posture is closer to a public archive than to a SaaS product.

IX

Where the trust comes from.

Trust does not come from claims, marks or branding. It comes from four structural properties, each carried on every record:

Structure. A record is recorded under a published schema. Its shape is the same for every contributor.

Provenance. Every record carries a canonical hash, a contributor identity, a locked timestamp and an evidence manifest with per-item hashes.

Permanence. Locked records are append-only. Corrections appear as visible addenda. The history remains readable.

Symmetry. Null and negative outcomes are preserved alongside positive ones, set in identical typography, and not coloured by interpretation.

The trust model is the product.

X

Governance principles.

The principles below are structural commitments. They are written into the schema, the contributor terms, and the platform itself. They are not marketing positions.

1. Neutral evidence infrastructure. Proof is not a manufacturer, a supplier, a certifier or a brand. It records and publishes evidence on behalf of contributors.
2. No product endorsement. Proof does not rank products, recommend products, or guarantee outcomes. No Proof Score. No outcome colour-coding.
3. No silent edits. A locked record cannot be quietly rewritten. Corrections appear as addenda and remain part of the ledger.
4. No farm data resale. Private envelope contents are not sold, shared or projected into commercial datasets.
5. Contributors do not pay to contribute. Free contributor access is part of the trust model, carried into the contributor terms.
6. No advertising on public surfaces. No paid placement on records, no paid product highlights in the Records Explorer, no sponsored insertion of records into reading flows. The public surface carries the records and nothing else.
7. Null and negative results stay. Removing them would make the system less trustworthy. They are preserved with the same weight as positive outcomes.
8. Commercial access does not buy deeper record visibility. Paid analytics surfaces sit above the public projection, never beneath it.
9. Conflicts are declared, not hidden. A declared conflict changes how a record is read. It does not invalidate the record.
10. The public projection is the canonical surface. Every public Proof URL is a typeset document, citable, sharable, and stable over time.

XI

Who Proof serves.

Proof’s records have one canonical surface but several reading audiences. Each audience reads the same record, in the same shape, for different reasons.

XII

Contributors do not pay to contribute.

Companies pay for analytical depth, attestation marks, infrastructure integration and institutional research partnerships. Individual contributors are never charged, regardless of their employer’s relationship with Proof.

Notably absent: contributor paywalls, freemium gates, data resale to farmers, advertising on public surfaces, paid product placements on records, paid removal of records.

XIII

Why now.

Four structural shifts make a permanent agricultural evidence layer more important than at any previous point.

Proof is built to close this gap from the evidence side, beginning with the categories where the gap is widest.

XIV

Roadmap.

Phases are described by what they make possible, not by quarterly dates. Each phase opens only when the density of records in the previous phase supports it.

XV

Make agricultural evidence inspectable at the scale of the work.

The long-term goal is straightforward. A permanent, public, inspectable evidence layer for every working farm. With the people and businesses behind the work protected by default. Built one locked record at a time, beginning with arable trials of biological inputs, and opening to the full breadth of agriculture as density grows.

Public evidence preserved. Private farm data protected. One record at a time.