- 01
The paper is a frozen answer to a question that keeps moving.
A journal article seals a result at the moment of acceptance, defends it against revision, and optimizes for a referee rather than a reader. By the time it prints, the data have grown, the world has turned, and the author has moved on. We mistook the artifact for the knowledge.
The crisis of uncited, unreplicated, unread papers long predates AI — a model writing hundreds of submission-ready manuscripts overnight only made the futility impossible to ignore. But the flaw was never the volume. It was the format.
- 02
So this one is built to be walked, not filed.
Climate Disclosure in Financial Statements still exists as a 75-page PDF behind a DOI. This is its other edition: the same inquiry rebuilt as a pipeline you can move through — from raw filing text, through the extraction of climate language out of the audited notes, to the aggregated evidence — read as a sequence rather than a sealed document.
The method is not buried beneath the conclusions; the method is the thing you can open and inspect. Each step is a tab you walk, not a paragraph you take on faith.
- 03
Living evidence, not a result sealed at acceptance.
The disclosure measures here are rendered from published aggregates and meant to keep moving: as new filings arrive, the panel grows and the same pipeline re-runs. The trend does not stop at the page a referee happened to see in 2024.
That is the seed of a research program rather than a paper — a long-running, AI-assisted inquiry that ingests evidence as it comes and holds its own claims up to continuous challenge. The PDF is the fossil. The pipeline is the organism.
- 04
One body of work, a different report for each reader.
A standard-setter, an investor, and a student ask different questions of the same evidence on whether financial statements see the climate. A frozen 75-page argument answers one of them, badly, and the rest not at all.
A program answers each in turn — a customized report drawn on demand from a shared base of evidence. This interactive edition is the first, simplest instance of that: the same analysis, re-cut for whoever is reading.
- 05
Authorship, as we know it, is ending — and that is not the loss it looks like.
The byline bundled three things: who did the work, who vouches for it, and who gets the credit. As tooling drafts, runs, and writes, that bundle comes apart. "Author" stops naming a craftsman and starts naming a position of responsibility.
What survives is not the name on the cover but the steward behind the program — the person accountable for the questions asked, the evidence admitted, and the claims allowed to stand. Less author, more guarantor.
- 06
Including this one.
The irony is worth stating plainly: an argument that the paper is ending is, for now, still a paper — four names, a DOI, a referee somewhere. You do not get to step cleanly out of a system to critique it.
This edition is one small defection from the format it argues against: not a document filed and forgotten, but a pipeline you can read, interrogate, and re-run. A first report, on demand, from a program that is only beginning.
The question is no longer how to publish more. It is what is worth knowing, who keeps that knowledge honest, and how a program — not a paper — is held to account.