SAVRN
Search Contact SAVRN

SAVRN Research · Editorial standard

AI infrastructure
research methodology.

How SAVRN sources evidence, labels claims and updates its research.

We separate facts, calculations, design assumptions, proposals, and operating evidence so a reader can tell what is known and what remains conditional.

Editorial owner
SAVRN Research
Methodology updated
SAVRN ResearchEvidence → claim
  1. Source record

    Identify the document, its date and the evidence it contains.

    SourceDateScope
  2. Claim & boundaries

    State what the evidence supports and what remains conditional.

    StatusUnitsAssumptions
  3. Review & update

    Keep the editorial context visible as the evidence changes.

    ReviewCorrections
From source to published claim

01 / Source hierarchy

Primary sources
and technical evidence.

We prefer primary records and technical documentation. Secondary reporting helps us discover developments and understand their context; material claims are checked against primary records where practical.

01 / Sources

Public records & decisions

Start with the record behind a policy, project or infrastructure decision. Read the document in its jurisdiction and at the stage it describes.

Examples

Statutes, agency records, permits, utility filings and system-operator filings.

02 / Sources

Research & technical documentation

Use technical sources to understand methods, equipment and operating requirements. Keep test conditions, specifications and limitations attached to the claim.

Examples

Research papers, standards and manufacturer documentation.

03 / Sources

Reporting & context

Use secondary reporting to discover developments and understand their context, then check material claims against primary records where practical.

Examples

News coverage and other secondary accounts that point to the underlying evidence.

Across every source type

Keep the comparison
and its limits visible.

  • Date and link material sources.
  • Preserve units and comparison boundaries.
  • Label estimates, assumptions and unresolved contradictions.
  • Avoid converting a reference design into an operating claim.

02 / Claim labels

Research claims
and evidence labels.

A published record, a calculation and a reference design establish different things. The label and its boundaries help readers assess the evidence without treating every statement as an operating result.

Fact supported by a source

What it means

A statement supported by an identified record, read with its date, scope and context.

What to check

What does the underlying record actually establish?

Calculation or estimate

What it means

A result derived from inputs and a method. Units, assumptions and comparison boundaries determine how to interpret it.

What to check

Are the inputs, units and method comparable?

Design assumption

What it means

A condition or target used to develop a reference design. It describes the basis of the design rather than measured operation.

What to check

Which conditions must hold for the design to apply?

Proposal or planned work

What it means

An announced, introduced or proposed action. Its status depends on the decisions, approvals and commitments still required.

What to check

What has happened, and what remains conditional?

Operating evidence

What it means

Records supporting what has been built, commissioned or measured in operation. A rendering or program model does not establish an operating result.

What to check

Which records support the specific operating claim?

Unresolved question

What it means

An evidence gap or contradiction that remains open. The uncertainty stays visible rather than becoming an unsupported conclusion.

What to check

What evidence is missing or in conflict?

03 / Editorial review

Editorial review
and research updates.

Review establishes the context for a claim. Updates keep that context useful when a source, project status or underlying assumption changes.

  1. Identify the source and scope.

    Connect the claim to its source, date, units and comparison boundary. Establish whether it describes a fact, calculation, assumption, proposal or operating result.

  2. Review the claim in context.

    Keep estimates and unresolved contradictions visible. Material topic pages carry a review date and editorial owner so readers can understand the review context.

  3. Update when evidence changes.

    Update the page and its source register as material evidence changes. A review date describes the editorial review; it does not certify every underlying project or forecast.

04 / Corrections

Submit a
research correction.

If a claim needs correcting or new evidence changes its context, send the page, the specific claim and the supporting record to SAVRN Research.

[email protected]

Include three things.

  1. The page URLLink to the SAVRN page you are referencing.
  2. The exact claimQuote or identify the statement that needs attention.
  3. The supporting sourceShare the record or evidence behind the correction.

Explore SAVRN research

Explore AI
infrastructure research.

Use the trackers, source registers and reference designs to examine AI infrastructure questions. The SAVRN Index also carries its own methodology and comparison standards.

SAVRN Index methodology

Read the separate standard behind the Index’s infrastructure research and comparisons.

Read the SAVRN Standard

For project discussions, explore how to work with SAVRN