Sources and Research Methodology

Source hierarchy

Sources are used in the following order of preference:

1. Primary institutional and scientific sources — peer-reviewed research in journals such as Nature and Science; official publications from NASA, ESA, NIST, WHO, the IEA, the OECD, the European Commission, UN agencies, national statistical offices and central banks.
2. Recognised research programmes and indices, such as the Stanford AI Index, where methodology is published.
3. Regulatory and standards documentation, including device authorisation registers and published technical standards.
4. Primary company documentation — technical papers, filings and specifications — used for claims about a company's own products, and treated as interested testimony rather than independent evidence.

Secondary reporting is used to locate primary material, not as a substitute for it. Where a claim cannot be traced to a source that meets this bar, it is either omitted or explicitly framed as uncertain.

The five categories of statement

Confusion between these categories is the most common failure in technology commentary, so each is used deliberately.

1. Current evidence

An observation supported by measurement or documented practice. Signalled by phrasing such as "current evidence suggests" or "as of publication". Cited.

2. Interpretation

The author's reading of what the evidence means — a judgement about significance, mechanism or direction. Attributable to the author, not to the source.

3. Forecast

A probabilistic statement about a specific outcome in a specific timeframe. Used sparingly, and never without stating the uncertainty attached to it.

4. Scenario

An internally consistent account of how conditions could combine. Scenarios are labelled and are not ranked as predictions. They exist to test assumptions, and a scenario set normally includes at least one uncomfortable outcome.

5. Speculation

A possibility with no established mechanism or evidence base. Always labelled as speculative.

Method for forward-looking analysis

• Separate capability (what has been demonstrated) from diffusion (what reaches ordinary use, and when).
• Apply base rates from comparable historical cases before accepting a claimed timeline.
• Reason from constraints — energy, materials, capital, skills, regulation, trust — which are more predictable than breakthroughs.
• Trace second-order effects, including rebound effects that run against the first-order intuition.
• Publish indicators that would confirm or falsify the analysis, so the reasoning can be checked later.

The full treatment of this method is set out in How to Think About the Future Without Pretending to Predict It.

Handling of numbers and dates

Figures are given with their source and, where relevant, the period they refer to. Where a widely repeated statistic cannot be traced to a primary source, it is not used — even when it is commonly cited elsewhere.

Limitations

This is independent single-author analysis. It has no peer review process, no institutional research budget and no proprietary data. Its value lies in synthesis, source discipline and explicit labelling of uncertainty, not in original empirical research.