Added#
trazum commitment — what a committed-use deal would have been worth on the
traffic you actually had. Chapter nine of docs/plan-1.51.md. Providers sell
these, and every team that signs one is doing arithmetic in a spreadsheet against
a number they guessed — the failure this product exists to end, at the highest
stakes it occurs, because the guess is annual and signed.
month list would pay saving
2026-01 $5,000 $4,000 +$1,000
2026-02 $5,000 $4,000 +$1,000
2026-03 $600.00 $3,000 -$2,400
2026-04 $4,000 $3,200 +$800.00
Net over 4 measured months: $400.00.
! 1 of them fell short, and the floor you would have paid for capacity nobody
used comes to $2,520.Both directions priced, because one direction is the sales pitch. A commitment is a floor as well as a discount. The deal above is net positive and one month cost $2,520 — netted together that disappears, and the disappearing is what a vendor's slide relies on. The unused floor is kept as its own figure and never folded into the saving.
An as-if calculation, and the wording never blurs it. "On the traffic you
actually had, this would have saved $X" is a measurement of the past; "you will
save $X" is a claim about the future, refused here as everywhere since 1.27.
Nothing is annualised, extrapolated or fitted to a trend, and the document
carries provenance: 'measured-past' for a machine reader.
The shortfall risk is a count of real months, never a probability. "Three of your last twelve would have fallen short" is a measurement. "There is a 25% chance of shortfall" is a model of a distribution nobody fitted, wearing the authority of arithmetic. Only the first is available from a log, so only the first is printed — with the measured spread beside it.
Partial months are dropped, not scaled. A fortnight replayed against a monthly floor is a shortfall the traffic never had.
Fewer than three whole months is a refusal, with how many more would settle it: a commitment is signed for a year, and an answer from one month is a year-long decision made on a fortnight of evidence. The break-even is stated anyway, because it is a fact about the deal rather than about the traffic.
A history shorter than the term still gets an answer, with the gap marked. Six months against a twelve-month deal is a real answer about six months; what it must not do is go unsaid.
Fixed#
The README's Action pins were advanced to the wrong commit. Reaching for the last merge SHA in hand gave the feature commit for chapter eight, labelled as if it were the release that follows it. The pin guard from 1.31 resolved the SHA, compared it against the version in the comment and failed by name three times. Every release since 1.28 has had two candidate commits here and the wrong one is always the more recent, which is what makes that guard load-bearing rather than decorative.
formatUsd rendered a value just under a thousand in the sub-thousand
format. 5000 - 5000 * 0.8 is 999.9999999999999, which took the two-decimal
branch and came out as $1000.00 — a string the thousands branch would never
produce, sitting in a column beside $5,000 and reading as a different currency
format for the same magnitude. The branch is chosen on the rounded value now,
so the boundary is the number a reader sees rather than the number the machine
holds. Found by looking at a real commitment table.
A signed column was using the unsigned formatter. formatUsd renders a
negative as $-2,400, which reads as a typo; in a column where every row can go
either way the sign carries the whole meaning, so it belongs in front of the
currency where a reader expects it. formatSignedUsd has existed since 1.30 for
exactly this and was not reached for.