Skip to content
BracketsightFree · no signup

Sources

Every rate, threshold and limit in the engine lives in a versioned JSON rules file with its citations. This page is generated from those files — if a figure appears anywhere on this site, its source is below. Ruleset cliffcheck-rules-2026.draft-1: 7 files carrying 21 citations between them.

A rules file carries six things, and the citation is only one of them

Each file states a ruleset version, an effective range so a rule that changes mid-year is a new dated file rather than an edit to code, a verification status, a plain-language note describing how the values are meant to be read, the values themselves, and its citations — each with a URL and the date someone last opened it. Nothing numeric is written into a TypeScript file. The formulas that consume these values are set out on methodology, and every change to any of them lands on the changelog.

Of the 21 citations below, 6 carry no verification date. That is not a formatting gap: it means the document at that URL has not been fetched and read by anyone here, so the citation records an intention rather than an act. Where a figure rests on a citation like that, the figure is verified elsewhere or it is not verified at all, and this page says which.

Only a primary source closes a gap

Statute, regulation, IRS guidance, a Federal Register notice, or a government dataset — those close a question. A secondary source may cross-check a value and may never close it, and a row resting on one stays labelled until the authoritative document has been read. Where a source could not be reached, the value is left exactly as it was and the gap is recorded rather than quietly improved. That discipline is why several citations below point at govinfo rather than the obvious host: the regulations site and the Federal Register site both redirect automated requests to a block page, so a successful response from either proves nothing, while govinfo serves authenticated GPO text.

What each status means

  • SAMPLE_DATA — illustrative figures written so the engine could be built and tested; there is nothing in them to verify.
  • UNVERIFIED — the machine-readable field has not been advanced, which does not always mean the values are wrong — see the note below.
  • VERIFIED_STATUTORY — read verbatim against the statute or regulation itself, from an authenticated government text.

The UNVERIFIED label deserves the qualification. A primary-source audit dated 15 August 2026 read the applicable-percentage table against the controlling revenue procedure and found every band boundary and every basis-point value to match exactly, and it read nine of eleven contribution limits and all fifty-one Medicaid flags the same way. It recommended the status field move. The field has not been moved, so it still reads UNVERIFIED here — this page publishes the machine-readable value rather than a friendlier summary of it. The figures that audit found wrong were corrected, and the repealed rule it found was removed; all of it is on the changelog.

Pre-launch status: files marked UNVERIFIED or SAMPLE_DATA carry placeholder or illustrative values pending primary-source verification. Bracketsight does not launch until every file reads verified. Benchmark Silver premiums are sample data for six counties today, so every premium, credit and clawback figure derived from them is illustrative rather than a quote — and no enrolled agent or CPA has yet reviewed the tax logic, which is a launch gate set out in the editorial policy.

The applicable-percentage table

applicable-percentage.2026.json · version aca-applicable-pct-2026.draft-1 · status UNVERIFIED · effective 1 Jan 2026 to 31 Dec 2026

Sets the share of income a household is expected to pay toward the benchmark plan, band by band, and carries the 400% eligibility ceiling as a multiple and a sentinel rather than a hard-coded constant. Every credit figure on the site passes through this file twice — once for eligibility, once for the contribution.

The federal poverty guidelines

fpl.2025.json · version hhs-fpl-2025.draft-1 · status UNVERIFIED · effective 15 Jan 2025 onward

The poverty line for a family size in each of the three state groups. 2026 coverage deliberately uses the 2025 guidelines, because eligibility runs on the guidelines in effect when open enrollment began. Multiplied by four, this file is the cliff.

Retirement and HSA contribution limits

contribution-limits.2026.json · version contribution-limits-2026.draft-1 · status UNVERIFIED · effective 1 Jan 2026 to 31 Dec 2026

Every ceiling the lever engine ranks against: elective deferrals and catch-ups, HSA limits by coverage tier, the traditional-IRA limit and its phase-out ranges, the SEP percentage and overall cap, and the self-employment tax constants used to compute the SEP base.

Medicaid expansion status by state

medicaid-expansion.2026.json · version medicaid-expansion-2026.draft-1 · status UNVERIFIED · effective 1 Jan 2026 onward

One boolean per jurisdiction, deciding whether a household below the poverty line is routed to Medicaid or told it is in the coverage gap. The flag is right and it is also the least expressive value in the engine — three states it cannot describe properly are named on the methodology page.

Cost-sharing reduction bands

csr-bands.json · version csr-bands.statutory · status VERIFIED_STATUTORY · effective 1 Jan 2014 onward

The three Silver-only cost-sharing bands and the 250% ledge at the top of them. Statutory rather than indexed, unamended since 2014, and it inherits premium-credit eligibility: no credit, no reduction.

Advance-credit repayment — the limitation, repealed for 2026

repayment-limits.2026.json · version aptc-repayment-limits-2026.repealed-1 · status VERIFIED_STATUTORY · effective 1 Jan 2026 to 31 Dec 2026

Whether a cap on repaying excess advance credit exists at all. For 2026 it does not, and the empty band array is the encoded rule rather than missing data. The reconciliation code throws if a future ruleset ever claims a cap while shipping no bands.

Benchmark Silver premiums

slcsp-sample.2026.json · version slcsp-sample-2026.draft-1 · status SAMPLE_DATA · effective 1 Jan 2026 to 31 Dec 2026

Two different things: a verified federal age-rating curve, and six invented county base premiums. The curve was diffed factor by factor against the CMS guidance with no mismatches. The premiums are illustrative and must be replaced, not checked.

Each of these is re-checked when its publisher publishes

The poverty guidelines arrive in mid-January and drive the following coverage year. HSA limits land around May. The §36B applicable-percentage table and the required contribution percentage arrive in mid-summer. Retirement and IRA cost-of-living figures come around October, the Social Security wage base around October and November, and the CMS marketplace public use files for the next plan year at roughly the same time. Medicaid expansion status is re-derived quarterly from the federal enrollment dataset rather than from a narrative page. Legislation amending §36B is watched continuously, because that is the failure mode this whole apparatus exists for: the audit above found a repayment cap that Congress had already repealed underneath a rules file nobody was watching. Indexed values drift on a schedule. Statutory structure does not.

Two known gaps will not close on a schedule. The 2026 Form 8962 and its instructions are unpublished until roughly January 2027, so the cliff convention rests on the 2025 and 2020 editions, which agree word for word across the ARPA boundary. And the federal list of state-specific age-rating variations was last updated in December 2021, with no newer version in existence. Both are described in full on methodology and in the about page's account of what is real data today and what is placeholder.

This is an independent estimate engine. It is not the Marketplace, not your insurer and not tax advice, and it cannot see your actual plan or your actual enrolment. Confirm any figure with healthcare.gov or your state exchange before you act on it. Nothing you enter leaves your browser.

Pre-launch build, and one limitation matters more than the rest. The second-lowest-cost Silver premium — the benchmark every credit on this site is calculated from — is sample data. The 6 county base premiums in the rule file are plausible invented figures, put there so the arithmetic could be built and tested; the real table is the annual CMS public-use landscape file, county by county, and it has not been ingested. The percentages, poverty guidelines, repayment limits and cost-sharing bands around it are verified against their primary sources, so the shape of the cliff and the direction of every lever are right — the dollar amounts are not yours until that table lands. The age curve the premiums are scaled by is the federal default and has been checked value by value. What is settled and what is not is set out on sources and methodology.