Democracy Monitor

Monitoring democratic institutions through public records

← Back to overview

What this site is — and is not

America is changing how it governs itself. Across the documentary record — executive orders, court filings, agency rules, inspector-general reports, congressional debate — you can watch authority moving: toward the presidency, past the referees, out of institutions that once checked it. This site exists to make that movement visible — through a lens we built, publish below, and correct in public.

We do not claim the shift is good or bad. The reasons for it are complex — some of it answers real frustrations with how government has worked — and where it leads cannot be known from here. Reasonable people disagree about whether the design America has carried for 250 years still serves the country. That debate belongs to voters, and this site does not take a side in it.

What we do claim is narrower, and we stake everything on it: the shift is happening, it is visible in the government's own records, and you can see exactly where. Every status on this site traces to specific public documents — quoted, linked, and machine-checked. Where the record shows a departure from long-standing practice, we say so plainly. Where it doesn't, we say that too.

What we decided before reading a single document

Every measurement on this site passes through choices we made first. We publish them so you can see through the lens — or around it.

  • Fourteen categories of institutional practice, chosen by us. A departure outside them is invisible here.
  • A page of historical norms, written by us, that says what "long-standing practice" means in each. The norms
  • Eight baseline years — every year of the two preceding administrations — so the same reviewer’s rates can be compared era to era. There is no baseline before 2017.
  • A two-pass reviewer. Two companies’ models, on purpose; neither model’s priors are inspectable by us, and the instructions each receives are public.
  • The words on the screen. We say departure; the instrument’s stored names — concern, erosion — predate this page and remain in every export.
  • Prose written by models, checked by code for its numbers and quotations before anything is published.

Each of these can be wrong; the reversals ledger records when they were. The full inventory — every model, every stored name, every prompt version — is its own page.

Why this stops at the record

Why keep watch at all, if we pass no judgment? Because some of what is changing took two and a half centuries to build and may not be rebuildable on any timeline that matters to the people alive now. A country is free to renovate its institutions. But renovation done quickly, without a record of what stood before, forecloses the option of changing course. This site is that record.

Keeping it is itself the act. Against a news cycle whose memory is measured in hours, the record is what stays. What to do about what it shows belongs to the reader — to voters, and to the question the Constitution assigns to Congress, not to a website.

How to catch us

Neutrality is a claim anyone can make. Ours is licensed by conduct:

  • The same instruments point at every administration. See the baselines
  • Our reviewer is tested by swapping the administrations’ names inside the documents it reads, and we publish what moves — including when it moves against us. See the swap audit
  • Its verdict rates are published for every era, side by side. See the rates
  • Fifty of its readings a quarter go to readers who are not us. The first fifty are ready, and we are still looking for the readers. Volunteer
  • The flags we raised under previous administrations are still published, unedited.
  • What this system cannot see is listed in our own words. Limitations
  • When we are wrong, we say so — on a page kept for that purpose. Reversals