Guides

What a guide is here

Each tool already carries a methodology page describing what its engine does, a sources page listing every rule file and the date its citation was last read, and a changelog recording what moved. A guide is none of those. It takes one borrower or one household, runs a single decision to the end, and says which way it comes out, what it costs to get wrong, and which part of it cannot be undone afterwards.

The bar a draft has to clear before it is published is that it contains at least one figure a reader could not have arrived at from the tool’s own pages — a crossover point, an eligibility interaction, a threshold two rules produce between them — and that it names the consequence rather than leaving the reader to infer it. A page that only restates what the calculator already prints does not get written.

So the collection grows by decision, never by keyword. Two guides chasing the same question split the traffic and start disagreeing within a year, so the content register in the repository holds one row per planned piece and refuses a repeated primary keyword. Where a tool has no guide yet, the list above says so instead of filling the gap.

Where a guide’s numbers come from

An article writes the name of a figure — loans.rap.topRate, aca.fpl.firstPerson — and the build hands back the value, the rule set version it was read from, the primary citation, and the date somebody last opened that citation and checked the value against it. The digits are never in the markdown. On a site about rules that move on a schedule that is the difference between a correction and a rewrite: the poverty guidelines are republished every January, the RISE rule took effect on 1 July 2026, and an article with a payment count typed into its prose would go on printing the old one long after the rule file had been fixed.

The same resolution runs for whole schedules — the RAP bracket table, the applicable percentage bands — so a guide cannot carry a table that has drifted from the one the engine computes with. A figure name that does not resolve fails the build rather than rendering a blank, which is why nothing here can be quietly worked around by typing the number in.

Values and schedules are what that engine has wired into the registry; open items are the ones whose source could not be closed. Nothing in the registry was last read off its primary source earlier than 8 Aug 2026, so that is the staleness floor for every figure a guide can print.

Property tax and trades sit at the bottom of that table, and that is the reason neither has a guide rather than an accident of scheduling. An article about a county appeal would have to state a filing deadline; the deadline is in the engine but not yet in the figure registry, so the article could only get it onto the page by typing it, which is the one thing this arrangement does not allow. Coverage first, then the article.

What these guides do not have

No named reviewer has checked either published guide. Both say so above the first paragraph, and the reviewedBy property is left out of their structured data entirely rather than filled with the site’s own name — a claim of human review is not a thing to assert by default. The authors page lists the six review gates the section editorial policies commit to and marks every one of them as not engaged. That is the largest open weakness on this site and prose cannot close it.

20 of the figures a guide can print carry an open item in the verification register — a value that is in the engine but rests on a source that could not be fetched, an agency table not yet published, or a documented simplification. Where a guide prints one, the register’s own words appear under the number, together with what would close the item, rather than a footnote a reader has to go looking for. The register itself is summarised on the about page.

Reading a guide next to its tool

Every guide ends in the tool it explains. For student loans the link carries the article’s worked example in the URL fragment, encoded by the calculator’s own encoder, so a reader lands on exactly the borrower the guide described — the same balance, rate, income and filing status — with all nine plans already ranked, and edits from there. A fragment is never sent to a server, and nothing about the scenario is stored.

The other four tools have no scenario transport: they seed from defaults and keep state in the browser, and none of them reads a query string. Their links are therefore plain links to the tool, and the component that renders them types the pre-fill option out of existence for those four. A link that looked pre-filled and then loaded somebody else’s defaults would show a reader an answer that was not the article’s.

Terms are defined in one place. A guide that needs discretionary income, the common level range or the benchmark plan links to its entry in the glossary, where every definition carries a primary source, rather than defining it again — two definitions of one term start disagreeing the first time one of them is edited.