The Uniform Lifetime Table, as a Maintained Artifact

A required minimum distribution is one division: balance over divisor. Everything that makes the answer defensible is in knowing exactly which divisor, from exactly which table, in exactly which publication year.

The required-minimum-distribution computation is famously simple and famously easy to get quietly wrong. The arithmetic is a single division — the prior year-end account balance divided by a life-expectancy divisor. The failure modes all live in the divisor: reading the wrong table, reading a superseded edition of the right table, or applying the right table at the wrong age. This piece is about how Worthune holds the table so that none of those failures can happen silently.

One table, sourced and bounded

The RMD model uses Table III of IRS Publication 590-B — the Uniform Lifetime Table, the one that applies to most account owners (those whose sole beneficiary is not a spouse more than ten years younger). The current edition is the set of tables effective since 2022, and the registry anchors it with sourced rows at its boundaries: the divisor at age 73 is 26.5, and the divisor at age 120 and over is 2.0, both cited to Publication 590-B's Appendix B. Between the anchors, the full run of divisors — age 73 through the terminal row — ships as a structured export the engine consumes directly, so the model and the registry cannot drift apart: they are reading the same object.

RMD = prior December 31 balance ÷ divisor(age).  At 73: balance ÷ 26.5.  At 75: balance ÷ 24.6.  At 120 and over: balance ÷ 2.0

The cohort question the model handles first

Before any divisor applies, there is the question of when RMDs begin, and this is where stale tools fail loudest. SECURE 2.0 moved the beginning age by birth cohort: the RMD model implements the current cohort schedule, and its spec documents the mapping. A tool built in the era of age 70½ — or updated once for the SECURE Act's 72 and never again — computes confident distributions for people who owe none yet. The model's approach is the registry's approach: the cohort rule is versioned behavior documented in the spec, so when Congress moves the age again, the change arrives as a spec version bump with a changelog entry rather than as a silent edit.

What the envelope shows

Run the RMD model through the API and the response's facts array cites the Uniform Lifetime Table rows — id, label, value, period, and source name — alongside the computed distribution. The citation is not decoration; it is the audit trail for the one number the computation depends on. A stored envelope from this year, retrieved in five years, will name the 2022-edition table it used, which is exactly the information a reviewer needs to confirm the answer was right when given even if the tables have been revised again since.

• ✓Confirm the table — Uniform Lifetime is the default; a spouse more than ten years younger uses a different one, and the model's spec says so.
• ✓Confirm the edition — divisors without an edition stamp are unverifiable.
• ✓Confirm the cohort — the beginning age depends on birth year under SECURE 2.0.
• ✓Use the prior December 31 balance — the divisor is only half the input story.
• ✓Read the envelope's facts array — that citation is the audit trail.

A table is the registry pattern at its clearest

Most registry rows are single constants; the Uniform Lifetime Table is forty-eight of them moving as one object with one source. That makes it the cleanest demonstration of what the registry is for: a government-published, occasionally-revised, edition-sensitive data structure that models must read exactly and that users deserve to see cited. The division is trivial; everything that defends it — table, edition, cohort, citation — is the craft.