Writing White Papers

The sequence that makes a finding verifiable

Drift is the normal condition for a paper written over weeks while data, code, and prose all change at different rates and on different days. A claim that no longer matches the number behind it. A figure regenerated from a different parameter set. A manuscript edited after the analysis it reports. Each of those is a silent divergence, and the repair for it is rarely sharper writing. What fixes it is a sequence where each stage checks the stage before it, so the divergence surfaces while the author still remembers the context that caused it.

This book builds that sequence. Seven chapters, seven artifacts, one repeatable pipeline that runs from a research question to an upload-ready PDF and can be pointed at the next paper without rebuilding from scratch.

Seven stages, seven working artifacts

• the-white-paper-pipeline — The full sequence as a working document: seven stages, what Claude does at each one, and the human checkpoints that sit between them. The structure of a paper that can be re-derived, not reconstructed.
• structuring-the-argument — An argument skeleton where every assertion carries a source tag. Downstream stages check against those tags — so the audit has something concrete to verify and the claim hierarchy is legible before the first word of prose is written.
• the-audit-loop — A script that parses the finished manuscript, recomputes every tagged claim straight from the source dataset, and exits nonzero the moment one number disagrees. The gate that runs between every revision and the next stage.
• reconciling-numbers — What to do when the audit catches a real failure: a directional reversal where the draft says “higher” and the data says “lower.” Tracing the discrepancy to its origin, deciding whether the draft was wrong or the data was misread, and revising so the two agree.
• submission-compliance — The preprint checklist, CARE checklist for case reports, an IRB exemption rationale, and the route to meeting medRxiv requirements for an author who holds no university affiliation. Each document parameterized to the study, not filled in generically.
• building-the-pdf — Stripping the source file, rewriting asset paths so figures resolve, injecting document metadata, and printing to PDF through a headless browser. The typeset output that matches what the preprint server will accept.
• polish-and-sign-off — Quality metrics applied to academic prose, followed by the sign-off checklist that clears the work before upload. A second paper costs a fraction of the first because the structure carries over and only the inputs change.

A preview

Sound data sat at one end. The path from there to the stated claim had been paved over as the work moved forward. The figure had been accurate when generated; reconstructing it a month later meant guessing which combination of script, parameters, and input files had produced it — and the guesses disagreed by enough to change the conclusion.

The reviewer’s question — where did this number come from — has a one-line answer when the work is built this way, and that answer is a file, a version, and a command.

Researchers turning findings into a defensible preprint

Anyone who has finished data collection and analysis and needs to move from results to upload-ready paper — without losing the thread between the numbers and the claims. The sequence handles both the writing and the verification in a single locked pipeline. It picks up where Research Navigator (Book 18) leaves off, at the point where a methodology exists and a dataset has been collected. The audit discipline it rests on extends the claim-verification approach from Self-Verifying Pipelines (Book 7).

A longer excerpt is available to newsletter subscribers.

Buy Ebook on Amazon