About

The Lagging Truth is a body of research about a single, under-noticed flaw in how the world measures itself. Many of the numbers institutions steer by are trailing averages — built entirely from the past, able only to follow what they measure, never to lead. Usually the lag is harmless. In certain high-stakes systems it is not, and the harm surfaces so late, and so far from its cause, that no one traces it back. The research maps where the lag is harmless and where it bites, and proposes fixes.

The work lives as ten research papers, published with their code, their sealed data, and their verification records; a plain-English companion for each paper, written so that a curious reader with no background can genuinely understand the research; and a set of dated, public predictions that put the framework’s claims on the record before their outcomes are known.

Behind the work is an independent researcher studying a systemic blind spot: from individuals to institutions, consequential decisions rest on stale information — and by the time things go wrong, the cause is long gone.

Two things about this work are stated plainly rather than discovered later. The research leaned heavily on AI: the author set the questions and made the decisions on record, while an AI system drafted the code, the experiments, and the prose under that direction — and each paper’s pre-publication review was itself performed by an AI in an isolated, sealed-package session. And the author is not a credentialed economist or mathematician. Both facts are exactly why the verification machinery exists: every claim is published with the code and data to check it, every paper carries a public corrections log, and the predictions page keeps score in the open. Experts are invited — genuinely — to test the claims and report what breaks.

The assembly line, the checklist, and the eye hospital

There is a precedent for work built this way. Early automobiles and aircraft were the objects of master craftsmen — machinists who held every part in memory, whose machines were exactly as good as their makers. The assembly line changed the question. Instead of asking where to find extraordinary people, it asked how to divide the work into steps that ordinary people could do superbly after a short training. Quality did not fall; it rose — because for the first time every step could be measured, checked, and fixed.

Aviation learned the same lesson the hard way. In 1935, Boeing’s Model 299 — the airplane that became the B-17 — crashed on its demonstration flight with one of the Army’s most experienced test pilots at the controls. The verdict of the day was that it was simply too much airplane for one man to fly. The fix was not a better pilot. It was a piece of paper: the pre-flight checklist, an admission that when complexity outgrows memory, process must catch what mastery drops.

The Aravind eye hospitals in India completed the argument. A retired surgeon decomposed cataract surgery into a repeatable sequence of stations and trained teams to run each one. The result performs surgery at enormous volume, most of it for patients who pay little or nothing, with measured outcomes that stand comparison with systems many times richer. Breaking the craft into steps did not cheapen it. It made excellence reproducible — and reproducibility is what made it available.

Science may be reaching the same point. Its process can be decomposed the way manufacturing and surgery were: questions registered before answers, data sealed and fingerprinted, code committed, every printed number tied to a script that regenerates it, review run in isolation, corrections logged in public, predictions dated before their outcomes. Each step is small, teachable, and checkable — and what AI changes is the length of the training program. With the steps written down and a capable assistant at every station, serious research stops requiring a rare polymath and starts requiring what it should have required all along: curiosity, honesty, and the willingness to be checked.

That shifts what counts as passing. Peer review asks whether a few busy experts find a result credible — a reasonable way to ration scarce attention, and no enemy of this project. Replication and forward prediction ask a harder question that anyone can grade: does the world agree? This site is built to be judged by the second standard, because the second standard is the one that scales to everyone.

It also changes who gets to look — and here this project does mean to point at something. Most people assume published science is publicly available. Mostly, it is not. The standard arrangement works like this: the public funds the research through taxes; scientists give their papers to journals for free; other scientists review them for free; and the finished work is then sold back — to university libraries at subscription prices that have strained even the wealthiest institutions, and to everyone else at tens of dollars per article — by publishers whose profit margins rival the most profitable companies in any industry. The newer “open access” model often just moves the tollbooth: readers no longer pay, but authors pay thousands per article to publish, which prices out precisely the independent and unfunded researchers the change was meant to include. Scientists dislike this arrangement more than anyone — the entire preprint culture exists to route around it — and policy has finally begun to shift. But almost nobody outside academia knows the arrangement exists at all; the author didn’t, until this project ran into it. It is, fittingly, the same species of flaw this research studies: a structural problem whose cost lands far from its cause, invisible until someone traces it. The research here is free by construction, permanently — the papers, the explainers, the code, the data recipes, the predictions, and the mistakes. Applications built on the research may someday carry a price; the research itself never will. That is the Aravind arrangement, adopted deliberately.

Whether work built this way can produce science worth trusting is itself a claim of the falsifiable kind. This site is one run of the experiment, conducted in public: the gates are published, the errors are logged where anyone can read them, and the predictions carry dates. Grade it.

The how it’s checked page explains the machinery in full.