​

Phase IV is the continuous evidence and learning function across every phase.

Phase IV — Continuous Operations does not wait for Phase III. It begins when Phase I creates the first usable baseline and continues for as long as the farm, its records, or its claims remain active. Its purpose is to keep the farm legible over time:

• what changed;
• what stayed stable;
• what failed;
• what was repaired;
• what was harvested;
• what was sold;
• what was forecast;
• what became actual;
• what evidence became stale;
• what risks materialized;
• what practices were implemented;
• what outcomes were measured;
• what claims need to be narrowed;
• what decisions should change next.

Time does not make evidence stronger by itself. Comparable, well-governed records over time make stronger analysis possible.

---

​

Phase IV at a glance

---

​

Phase IV connects four different things

Phase IV is strongest when these are kept distinct.

Phase IV is useful only if evidence can change decisions.

---

​

The Phase IV operating model

​

1. Capture the right record

Not every farm event deserves the same record. Useful record families can include:

• farm activity;
• planting / establishment;
• harvest;
• sale / distribution;
• expense / procurement;
• labor / role;
• field note;
• soil sample;
• water measurement;
• vegetation observation;
• biodiversity / survival;
• infrastructure / maintenance;
• incident / risk;
• training / competency;
• milestone / acceptance;
• partner / buyer interaction;
• public-goods delivery;
• certification / compliance status;
• governance / funding decision.

The record should fit the question.

---

​

2. Preserve evidence type

Phase IV should keep the Framework’s evidence vocabulary visible. Do not let a dashboard or report silently convert one type into another.

Forecast ≠ actual. Observation ≠ measurement. Measurement ≠ causality. Attestation ≠ certification. Interpretation ≠ raw evidence.

---

​

3. Use a governed record lifecycle

Kokonut Intelligence currently distinguishes record lifecycle states such as:

draft → submitted → verified / rejected → published

That lifecycle is useful for workflow control. But the words must be interpreted carefully.

​

Verified

Within Kokonut Intelligence, Verified means the record passed the applicable internal review workflow. It does not automatically mean:

• independently audited;
• scientifically validated;
• regulator-approved;
• organically certified;
• carbon-verified;
• legally adjudicated.

​

Published

Published means the record is available to the intended downstream/public surface. It does not mean the record is:

• complete forever;
• still current;
• free of uncertainty;
• appropriate for every later claim.

​

Rejected

A rejected record can be valuable evidence about:

• missing provenance;
• inconsistent units;
• unsupported claims;
• duplicate data;
• invalid method;
• insufficient review.

Phase IV should preserve rejection/quality history rather than hiding it.

---

​

4. Keep freshness visible

A high-quality record can become stale. Every material current-state claim should make it possible to answer:

• when was the source event?
• when was the record created?
• when was it reviewed?
• when was it published?
• when was it last updated?
• is a newer record expected?
• does the claim still describe current farm state?

​

Example

“The farm has seven jobs.”

That can be a valid documented reference. For a current employment claim, Phase IV should still ask:

• seven as of when?
• full-time, part-time, seasonal, or role count?
• are the roles still active?
• what changed after the source period?

Publication status and freshness are different properties.

---

​

5. Compare like with like

Longitudinal analysis requires compatibility. Before comparing records across time, check:

• same farm / plot / asset boundary;
• same unit;
• same method;
• same instrument or documented instrument change;
• same or comparable timing/season;
• same denominator;
• same definition;
• same lifecycle semantics;
• known missing-data periods;
• known intervention changes.

​

Method changes should be explicit

If a method changes:

1. record the old method;
2. record the new method;
3. state the change date;
4. explain whether values remain comparable;
5. avoid creating a fake continuous trend when comparability was broken.

---

​

6. Replace forecasts with actuals — without deleting the forecast

Phase IV should preserve the original assumption and the later result. That enables learning. Examples include:

• planting quantity;
• survival;
• harvest timing;
• yield;
• loss;
• saleable quantity;
• realized price;
• buyer/channel;
• labor;
• input use;
• cost;
• maintenance;
• public-goods transfer/delivery;
• project milestone.

A revised forecast should not erase the evidence that the earlier forecast was wrong.

Forecast error is part of the farm’s learning history.

---

​

7. Maintain the farm — not only the dataset

Continuous operations include physical maintenance. Phase IV should keep current records for assets such as:

• irrigation;
• pumps;
• tanks / reservoirs;
• nursery;
• biofactory;
• poultry infrastructure;
• tools;
• fencing;
• buildings;
• energy systems;
• monitoring devices;
• storage;
• access roads.

Useful maintenance evidence can include:

• condition;
• service date;
• downtime;
• repair;
• replacement;
• spare parts;
• vendor;
• cost;
• safety issue;
• unresolved defect.

---

​

8. Update the risk register from real events

Phase I starts with risks. Phase II reveals which ones materialize. Phase III builds stronger controls. Phase IV keeps the risk model current. Examples:

• drought;
• storm/flood;
• disease/pest;
• crop mortality;
• water shortage;
• pump failure;
• buyer loss;
• price decline;
• labor shortage;
• key-person dependency;
• safety incident;
• certification correction;
• legal/land change;
• data gap;
• sensor failure;
• privacy issue;
• methodology weakness;
• funding delay.

For each material risk, update:

• evidence;
• likelihood / uncertainty;
• consequence;
• owner;
• mitigation;
• trigger;
• residual risk;
• status;
• review date.

A risk register that never changes after real incidents is not functioning as a learning tool.

---

​

Evidence quality matters more than elapsed time

The old Phase IV page used a time ladder:

one month = observations → one year = baseline → three years = patterns → five years = compounding credibility

That is too deterministic. A baseline can exist on day one if it was collected properly. Five years of inconsistent or stale records can still be weak evidence.

​

Stronger longitudinal evidence depends on

---

​

Cadence should follow the record and claim

There is no universal Phase IV calendar.

Use annual reporting when annual reporting serves the claim or stakeholder—not because Phase IV requires one report every year.

---

​

Publication, privacy, and provenance

​

Not every record should be public

Phase IV should distinguish:

• public evidence;
• partner/funder evidence;
• internal operational evidence;
• restricted personal/commercial/legal evidence.

Publication decisions should consider:

• personal information;
• worker privacy;
• exact locations;
• security;
• buyer/vendor confidentiality;
• legal agreements;
• financial sensitivity;
• conservation sensitivity;
• consent;
• public-interest value.

​

Public evidence should be deliberately selected

A public record should answer:

• why is this useful publicly?
• what claim does it support?
• is sensitive information removed?
• is the source/method clear?
• is the reporting period clear?
• is the record current enough?
• is the interpretation bounded?

---

​

Kokonut Hub

The Kokonut Hub is the public surface for selected published project/farm information. Use it for:

• current published farm records;
• changeable project state;
• public evidence pointers;
• published operational/impact context.

Do not assume:

Hub = complete farm database

Some legitimate evidence can remain private, restricted, or unpublished.

---

​

EAS on Celo

Kokonut Intelligence can anchor selected evidence/claims through EAS on Celo. An attestation can help preserve:

• signer;
• timestamp;
• schema;
• evidence reference;
• provenance context.

It cannot make weak evidence strong. It cannot independently prove:

• ecological causality;
• carbon sequestration;
• biodiversity restoration;
• certification;
• profitability;
• legal rights;
• work quality.

Attest selectively when durable public provenance adds value.

---

​

IPFS / Filecoin or other durable storage

Content-addressed / durable storage can be useful for:

• reports;
• evidence artifacts;
• datasets;
• media;
• snapshots.

It is not a universal Phase IV requirement. Integrity and persistence are different from scientific validity.

---

​

Phase IV, MRV, Kokonut Intelligence, and the Hub

These layers should not be collapsed. This distinction matters because:

• a software workflow is not the methodology;
• a public page is not the entire evidence base;
• an attestation is not a scientific conclusion;
• an internally verified record is not external certification.

---

​

Phase IV feedback should go to the right authority

Phase IV evidence can inform many decisions. It should not route every decision to the DAO.

---

​

Phase IV and AI / agent workflows

Agents can make continuous evidence more usable. They can help:

• detect missing records;
• flag stale data;
• compare forecast vs. actual;
• identify anomalies;
• draft reports;
• assemble evidence packets;
• summarize risk changes;
• calculate defined indicators;
• check schema consistency;
• propose follow-up questions;
• prepare transaction or attestation drafts within authorized workflows.

They should not autonomously:

• certify ecological outcomes;
• approve farm funding;
• decide land rights;
• change farm operations without operator authority;
• invent missing data;
• convert an estimate into an actual;
• treat internal Verified as scientific truth;
• publish sensitive evidence without applicable permission;
• make unsupported causal claims.

Agent confidence is not evidence quality.

---

​

Phase IV across the development phases

Phase IV is therefore not a destination. It is the memory and feedback function running across the development lifecycle.

---

​

Adelphi — current Phase IV reference

Adelphi is currently the strongest live reference for Phase IV because it already combines:

• a documented Phase I history;
• active Phase II operations;
• ongoing MRV;
• public Kokonut Hub records;
• selected Celo EAS provenance capability through Kokonut Intelligence;
• remote-sensing / field-record capability;
• forecast models that need actual reconciliation;
• soil / vegetation / biodiversity evidence needs;
• infrastructure that requires maintenance;
• market and certification pathways that can change over time.

​

The most important Phase IV task at Adelphi

It is not simply:

collect more data

It is:

keep replacing static assumptions with current, comparable, reviewed records while making uncertainty and stale information visible.

Examples:

---

​

Replication learning

A farm’s long history can help the next farm. It still does not prove transferability. Useful replication evidence includes:

• which fields remained useful;
• which definitions needed adaptation;
• which crop assumptions transferred;
• which failed;
• which practices were locally appropriate;
• what operator capacity was required;
• what infrastructure was reusable;
• what market assumptions changed;
• what data collection was too expensive or unnecessary;
• what governance boundaries held;
• what technical integrations remained compatible.

​

Terra Viva

Terra Viva remains a Phase I planning-stage replication test. Phase IV evidence from Adelphi can inform Terra Viva’s planning. It should not be copied as if:

• crops;
• yields;
• prices;
• costs;
• soil response;
• biodiversity response;
• labor;
• buyers;
• public-goods economics;

will be identical.

Replication learning transfers questions, structures, and lessons more safely than it transfers conclusions.

---

​

External pathways: evidence enables diligence, not entitlement

Longitudinal evidence can make a farm easier for an external party to assess. That does not create an automatic capital pathway.

Phase IV can improve diligence readiness. It does not create a right to financing, certification, a credit, a premium, or tokenization.

---

​

Reporting should be decision-specific

A useful Phase IV report should identify:

• decision / audience;
• farm / boundary;
• reporting period;
• sources;
• actuals vs. forecasts;
• methods;
• material changes;
• missing data;
• stale records;
• incidents;
• risks;
• outcomes;
• uncertainty;
• relevant evidence pointers;
• optional attestations;
• next actions.

​

Examples

Not every farm needs all of these.

---

​

Common failure modes

---

​

Phase IV reviewer checklist

Before using longitudinal farm evidence for a decision or public claim, ask:

• What decision or claim is this evidence supporting?
• Is the farm/plot/population/system boundary clear?
• Is the record an observation, measurement, derived indicator, forecast, actual, attestation, or interpretation?
• Are forecasts separated from actuals?
• Is the source/provenance inspectable?
• Is the record current enough for the claim?
• Are lifecycle states understood correctly?
• Is internal Verified being mistaken for external certification?
• Are methods/units comparable across periods?
• Are method changes documented?
• Are missing-data periods visible?
• Are privacy/sensitivity constraints respected?
• Is publication necessary and authorized?
• Is EAS/IPFS being used only where useful?
• Are ecological claims separated from implemented practices?
• Are causal claims supported by the methodology?
• Are maintenance and incidents included where they affect the conclusion?
• Are risk updates current?
• Is the relevant domain reviewer involved?
• Is the decision routed to the correct authority?
• Are agent-generated outputs traceable to source records and reviewed?
• Is replication being inferred beyond the available multi-farm evidence?
• Is the next evidence/action explicit?

---

​

How different readers should use Phase IV

---

​

Current maturity

---

​

Continue through the Framework

Phase IV succeeds when the farm remembers accurately enough to learn: current records stay distinguishable from stale ones, actuals stay distinguishable from forecasts, evidence stays distinguishable from interpretation, and every important decision can be traced back to what the farm actually observed.