The se- rule for thousands is missing
their bug"satu ribu" instead of "seribu" — exactly the error PRD.md §2 names.
Published packages run against the same inputs. Factual, versioned, dated — a map of a shared problem, not a scoreboard.
Recorded 2026-08-17v25.8.1
Organised by failure mode, not by package — one place an implementation breaks can show up across several inputs and several packages at once.
"satu ribu" instead of "seribu" — exactly the error PRD.md §2 names.
Returns an ordinary placeholder string ("terlalu besar / not implemented yet"), not a structured refusal — a string that can reach a printed document undetected.
No word at all for an input of 0 — not a capitalisation difference.
A middle group and the final digit both change on the 2⁵³+1 input — consistent with number arithmetic rather than BigInt.
"Kuadriliun" follows KBBI’s standard adaptation of Latin qu- to ku- (kuadrat, kuadran); "quadriliun" does not, and is uncited anywhere — unlike the milyar/miliar and triliun/trilyun pairs that are recorded as variants.
Per-word Title Case over words and se- rules identical to ours. Document capitalisation is a cited option in lib/options.ts.
KBBI takes miliar; milyar is common in the field. Both sides are cited as an option (`ejaanMiliar`).
The precise reference behind the summary above — row by row, package by package.
Differences are marked, then classified in writing as our bug, their bug, or a legitimate variant — the record lives in tests/survey/.