Releases 1.99.0 – 1.138.0

Newest first. Three numbers, and which of them moves says what happened to you rather than how much work it was.

The first moves when something that worked stops working: a query that used to answer and now does not, a saved list that will not load, a link you shared that no longer opens what it opened. It has never moved, and that is the whole of what it is for.

The second moves when the page answers something it could not answer before, or answers it better than it did. New grammar is the usual reason. A correction is the other, and a correction is not a break even though the answer changes underneath somebody: "days be christmas" used to say 154 days and now says nothing, because it was reading Belgium out of the word "be", and taking that away left the page more right rather than less useful. An answer that gains a drawing of what it just said counts here too.

The third moves when nothing you type comes back any different.

Numbers get skipped when it takes more than one commit to introduce a release, so only the release from the last commit goes out.

1.138.031 July 2026

Fixed: a line of ordinals read as a date

1st 2nd 3rdSat, 1 Feb 2003nothing
1st 2ndSun, 1 Feb 2026nothing
Every written date marks at most one of its parts: "the 3rd of May 2027" marks the day and leaves the month and the year alone. A second mark says these numbers were never a date.
A wrong date stated confidently is the one outcome this engine exists to avoid, since nothing about it looks different from a right one.

1.137.031 July 2026

The operators that work on the bits

12 xor 10nothing6
5 << 25 Feb 202620
0xff & 0x0f12 Oct 202615
A number written in another base has read since the bases went in, and there was nothing to do to one. The reason anybody looks at 255 as 11111111 is to mask part of it off, shift it along or flip a bit in it.
Two of these lines were being read as dates, which is worse than answering nothing: a shift is not a day in February.
The bits are shown, since an answer in decimal to a question about bits has been converted back out of the notation that shows what happened.
"and" and "or" on their own stay out: they are the busiest joining words here, and a reader owning them would reach into every range and every list on the page.

1.136.031 July 2026

Fixed: a span rounded to the hour lost the half hour

hours between 9am and 5:30pm9 hours8 hours 30 minutes
hours between 9am and 5:15pm8 hours8 hours 15 minutes
A working day with half an hour quietly added to it, and the true figure was sitting underneath in minutes, so the answer contradicted its own footnote. The headline is the half anybody reads.
The unit asked for still leads, since that is what asking for hours means. What was under it is said in the next unit down instead of vanishing.
Seconds still round, having nothing smaller to be said in, which is the one place the rounding is not a choice.

1.135.031 July 2026

Whether the clocks have moved yet

is it dst in new yorknothingyes · summer time · UTC-4
is it daylight savings in sydneynothingno · standard time · UTC+10
does tokyo observe dstnothingno
When the clocks next move has answered since the clock changes went in, which is the question about the future. The question about right now is the one that gets asked.
An offset is half a fact without it: a meeting written down in January as "3pm New York, UTC-5" is at a different hour in July, and which of the two a place is keeping today settles whether the note still means anything.
The larger of a zone's two offsets is its summer one in both hemispheres, so the southern half needs no rule of its own.

1.134.031 July 2026

A mark out of a total

42 out of 60nothing70%
42 out of 60 as a percentnothing70%
The three orders this question already answered all carry the word "percent", which is the shape the line has when somebody already knows what they want.
A mark on a test, a poll, a count of anything against a total is written "42 out of 60" and wants a percentage without having said so, that being the only reason to write two numbers down that way round.
The mark itself goes underneath, since a percentage on its own has lost the two numbers it came from.

1.133.031 July 2026

A length of time as a stopwatch prints it, and two of them added

1h 30m in minutesnothing90 minutes
5m30s in secondsnothing330 seconds
2h 30m + 45mnothing3 hours 15 minutes
"1 hour 30 minutes in minutes" has read since the compound quantities went in, and the short form had nowhere to go. "m" is metres everywhere else here, and the unit in front of it settles it: nobody writes down an hour and thirty metres.
The parts may run together, since the letters do the work the spaces would have done.
Two lengths added is the sum done off a stopwatch or a timesheet, and it had no reader at all: this could move a clock by an hour and a half and could not add one to another.

1.132.031 July 2026

Whether a number is odd or even

is 12 evennothingyes
is 17 evennothingno, so it is odd
is 12 odd or evennothingeven
The plainest thing anybody asks about a whole number, and the one thing this could not say about one, while answering about that same number whether it is prime, square, triangular, perfect or abundant.
A no says what the number is instead, since here the two answers are the whole of the question: "17 is not even" and "17 is odd" are the same sentence and only one of them has told you anything.
Asked as the choice it is, too, which is how it gets typed by somebody who does not have one of the two words in mind already.

1.131.031 July 2026

10C3, the way a textbook prints it

10C3nothing120
10P3nothing720
nCr 10 3nothing120
"10 choose 3" is how it is said and 10C3 is how it is written, and this read only the saying of it. A textbook prints the welded form and a calculator has nCr on a key, so both of the shapes anybody copies were missing.
The two numbers may now be separated by a space alone once the name is in front of them, since a comma between two things that are plainly two things is punctuation nobody types. "gcd 12 18" comes along with it.

1.130.031 July 2026

How long until a time of day

hours until 5pmnothingin 21 hours
minutes until 3pmnothingin 1,140 minutes
hours until noonnothingin 16 hours
The span reader wants both ends, which is right for the length of an afternoon and wrong for the afternoon you are sitting in: nobody types the time it is now.
Midnight has been countable since the day engine learned clock units, and it is the only time of day that engine can hold, a day boundary being the only clock reading a day-granular thing carries. Every other hour had nothing to be counted to.
A time already gone today is that time tomorrow, and the answer says which. Midnight stays where it was, so no line has two readers.

1.129.031 July 2026

The clock takes the sign spelled out

3pm plus 90 minutesnothing4:30 pm
3pm minus 90 minutesnothing1:30 pm
now plus 3 hoursnothing11:00 pm
The sums have read "plus" and "minus" as words since the day they went in. The clock read only the two marks, so the same sentence answered with a sign and stayed silent with a word.
Turned into the marks in one place rather than taught to the pattern, which reads them twice and would have wanted teaching twice.

1.128.031 July 2026

What the date is

what is the datenothingFri, 24 Jul 2026
todays datenothingFri, 24 Jul 2026
what is the date in tokyonothingFri, 24 Jul 2026 in Tokyo
The plainest thing a calendar can be asked, and one of the few this could not answer at all.
"today" has always read, and it answers with the word "today", which is true and is not what was asked: somebody typing this wants the date spelled out, because they are about to write it down on something.
A place may follow, since for a third of every day the date somewhere else is a different date, and that is the only reason to ask about anywhere but here.

1.127.031 July 2026

Which week a date falls in, and which quarter

what week is 25 dec 2026nothingWeek 52 of 2026
what quarter is christmasnothingQ4 2026
which day of the year is christmasnothingDay 359 of 365
"what week is it" has answered since the periods went in, which is the wrong half to have on its own: today is the one day a reader can find out by looking at anything to hand.
The week a meeting falls in and the quarter a deadline lands in are what actually gets looked up, and both are about a date somebody is holding.
Anything the date grammar reads stands on the right, so a feast or a weekday works as readily as a written date, and an impossible date is still reported as one.

1.126.031 July 2026

Trigonometry with the angle in degrees

sin 30 degreesnothing0.5 · degrees
cos 60 degnothing0.5 · degrees
tan(45°)nothing1 · degrees
The sums answer in radians and have always said so, which is right for a calculator and no use to somebody holding a protractor. A line that names its unit outright got nothing at all.
The conversion is written into the sum rather than done quietly, so what the parser adds up is what a reader would have typed themselves, and the echo can tell the two units apart by looking at it.
The inverse functions are left alone, since their argument is a ratio and it is their answer that is an angle.

1.125.031 July 2026

Per cent, spelled out

12.5 percent of 340nothing42.5
15 percent off 200nothing170
1200 + 15 percentnothing1,380
50 percent as a decimalnothing0.5
Seven rules here read the sign and not one of them read the word, which is the wrong way round: the sign gets typed by somebody with the key in mind, and the word by somebody saying the sentence.
The word becomes the sign once, so every one of those rules gets it at the same moment instead of growing a spelled-out twin. "per cent" and "pct" come along with it.
Not in front of "points", since a rise of five percentage points is not five per cent and telling those two apart is the whole reason that answer exists.

1.124.031 July 2026

The name of the sum, said after the numbers

1 2 3 4 and 10 24 30 42 correlationnothing0.989778266557
1 2 3 averagenothing2
1 2 3 4 mediannothing2.5
Every list reader wanted its word in front, which is the order a function is written down in rather than the order a question arrives in: you type what you have, and then what you want done with it.
Both readers now see the line with any trailing name moved to the front, so the word order is settled in one place and the two of them cannot disagree about it.
Two numbers at least, since one number and a short word is something else nearly every time: "90 min" is still an hour and a half.

1.123.031 July 2026

The chips advertise the grammar that arrived this week

what week is it(never offered)a chip
when is sunset earliest in vienna(never offered)a chip
midpoint between karachi and vancouver(never offered)a chip
how old will i be in 2044 if born 12 mar 1991(never offered)a chip
Fifty-eight things were added and not one of them was on a chip, so the row still advertised the calculator as it was a week ago. A grammar nobody can discover is a grammar nobody types.
Five new groups, since the sun, the holidays, the period we are in, ages and unix time each answer several questions now and had at most one chip between them.
Twenty groups where there were fifteen, and thirteen chips shown, so the row is different on every visit instead of being the same list with two swapped.

1.122.031 July 2026

What year it is in another calendar

what year is it in the hebrew calendarnothing5786
what year is it in the islamic calendarnothing1448 AH
what year is it in the persian calendarnothing1405 AP
japanese yearnothing8 Reiwa
The whole date has been readable since the calendars went in, and the year on its own is the part anybody asks for out loud: a year is what a calendar is identified by.
The whole date goes underneath, since a year in a calendar most readers do not use is a number with nothing to hold on to.
The number comes from the same formatter the full date does, so there is one place a calendar is read and no chance of two of them disagreeing.

1.121.031 July 2026

Saying why a day is not here

when is chinese new yearnothingLunar New Year isn't in here: it is set by a new moon rather than by a rule…
when is diwalinothingDiwali isn't in here: it falls on a lunisolar month whose length is observed rather than fixed…
The reasoning was written down beside the calendars and never said out loud, so the answer was silence, and silence reads as ignorance.
It looked like a calculator that has never heard of the largest annual holiday on earth, rather than one that has thought about it and refused.
Every other refusal here says why. These are the only ones where the reason is a judgement rather than a bound, which makes saying it more important and not less.

1.120.031 July 2026

A year named, in Roman numerals

next year in roman numeralsnothingMMXXVII
this year in romannothingMMXXVI
christmas in roman numeralsnothingMMXXVI
Roman numerals are for years nine times out of ten, and every one of those questions is about a year somebody would rather name than count out.
The left side is handed to the date grammar, which has known all of those since it was written.

1.119.031 July 2026

Two bases at once

255 in hex and binarynothingff and 11111111
255 in hex and octalnothingff and 377
10 in binary and hexnothing1010 and a
The same courtesy the units and the currencies were shown, and worth more here: a number looked at in one base is usually being looked at in the others too, that being what a base is for.
The same splitter and the same both-or-neither rule, so one base beside one refusal never happens.

1.118.031 July 2026

Hex of 255, said that way round

hex of 255nothingff
binary of 10nothing1010
base 16 of 255nothingff
"255 in hex" is a conversion written out. This is how the same thing is said when the base is the thing being asked for, and it is the order every other "give me X of Y" question on this page takes.
Loose on purpose and safe because of what it is checked against: the first word has to name a base, so "sum of 1 2 3" and "factors of 360" come nowhere near it.

1.117.031 July 2026

Unix time for a date, said that way round

unix time for christmasnothing1798156800
timestamp of christmasnothing1798156800
epoch time for 25 dec 2026nothing1798156800
The date had to come first, which is the order a conversion is written in and not the order a question is asked in.
"unix time for christmas" puts the thing wanted at the front, the way "distance from tokyo to london" and "sunrise in london" both do.

1.116.031 July 2026

Seconds since the epoch, and milliseconds

seconds since epochnothing1784851200
milliseconds since epochnothing1784851200000
unix time in millisecondsnothing1784851200000
"unix time" is what a programmer calls it and answered already. "seconds since the epoch" is what it is, and is how everybody who has to explain it writes it down.
The milliseconds were on the detail line all along and there was no way to ask for them.

1.115.031 July 2026

Seconds, which were the one clock unit not counted in

seconds until midnightnothingin 100,800 seconds
seconds until christmasnothingin 13,323,600 seconds
seconds in a daynothing86,400 seconds
Hours and minutes have driven a count since the clock units went in. Seconds did not, though a countdown is the one thing anybody counts in them.
Not the bare "s": it is already noise, being what is left of "what's" once the apostrophe goes, and a driver owning it would take the front off every question asked that way.

1.114.031 July 2026

Turning an age, and an age at a date

when do I turn 40 if born 5 jun 1990nothingWed, 5 Jun 2030
how old will I be in 2050 if born 5 jun 1990nothing59 or 60
how old will I be on 1 jan 2050 if born 5 jun 1990nothing59 years 6 months 27 days
"born 5 jun 1990" answers how old somebody is now, which is one instant of a question people ask about every instant of their lives.
A year named on its own is two ages, and saying so is the point: somebody born in June is 59 for half of 2050 and 60 for the rest, and one number would be wrong half the time without ever looking wrong.

1.113.031 July 2026

Anniversaries

10th anniversary of 5 jun 2020nothingWed, 5 Jun 2030
50th anniversary of 20 jul 1969nothingSat, 20 Jul 2019
1st anniversary of christmasnothingSat, 25 Dec 2027
The offset grammar has added ten years to a date since it was written, and "5 jun 2020 + 10y" is how it had to be asked. That is the sum rather than the question.
A 29 February anniversary lands on the 28th in a year without one, and the answer says so rather than leaving the reader to notice the day changed.

1.112.031 July 2026

Leap years counted over a span

leap years between 2000 and 2100nothing25 leap years
how many leap years since 2000nothing7 leap years
leap years in the 21st centurynothing24 leap years
One at a time is one answer; the run is another, and it is the one the century rule is interesting for.
Between 2000 and 2100 there are 25, and the arithmetic anybody does in their head says 26, because 2100 is not one.
Where the list is too long to read, the answer names the centuries taken out instead, since those are the whole reason the count is not the division you expect.

1.111.031 July 2026

The next leap year

next leap yearnothing2028 · 2 years away
previous leap yearnothing2024
next leap year after 2096nothing2104 · 8 years away
The rule could only ever be asked about a year somebody already had in mind, which is the backwards half of the question: the reason anybody looks a leap year up is to find out when the next one is.
Walked rather than worked out. The longest gap the Gregorian rule allows is eight years, 2096 to 2104, so eight tests settle it and the walk cannot disagree with the rule.

1.110.031 July 2026

A quantity with nothing said to convert it into

5 kmnothing3.10686 mi · 5,000 m · 16,404.2 ft
180 lbnothing81.6466 kg
20 cnothing68 f
convert 5 kmnothing3.10686 mi
The shortest thing anybody types at a converter, and it answered nothing, which reads as "I do not know that unit" when the truth is "I do not know what you want".
What they want is nearly always the other system, that being why anybody converts a length at all, so each dimension names the handful of units its questions get asked in.
Time is deliberately absent: "3 weeks" is a date offset, and a converter claiming it would answer 504 hours to somebody asking what the date is.

1.109.031 July 2026

Half a kilo

half a kilo in poundsnothing1.10231 pounds
a third of a mile in metresnothing536.448 metres
quarter of an hour in minutesnothing15 minutes
an eighth of a mile in feetnothing660 feet
A fraction in front of a size has been read since the sizes went in, and a fraction in front of a unit was not read at all, though it is the commoner sentence by a distance.
Nobody says "0.5 kg". They say half a kilo, and a quarter of an hour, and a third of a mile.
Only in front of something this file can convert, so "half a mind" is still not a quantity.

1.108.031 July 2026

Two currencies at once

100 usd in eur and gbpeur and gbp isn't a currency I have a rate for87.90 EUR and 75.05 GBP
100 usd in eur, gbpnothing87.90 EUR and 75.05 GBP
The same reading two units already got, handed to the currency reader instead. Which one takes the line is decided by which one answers, as everywhere else here.
The rate line is said once rather than twice, since both answers share their source and its date and an echo that repeats them reads as though the work had been done twice.
Before the currency reader rather than after, since one handed "eur and gbp" reports no such currency instead of noticing there are two of them.

1.107.031 July 2026

One quantity, two targets

20 c in f and knothing68 f and 293.15 k
100 km in miles and feetnothing62.1371 miles and 328,084 feet
1 kg in lb and oznothing2.20462 lb and 35.274 oz
A converter that takes one target makes you ask twice, and the second asking is the same line with three characters changed.
It costs no new arithmetic: the line is split at the "and" and each half handed to the reader that already answers it, so every unit and spelling comes along unchanged.
Both or neither, since one good answer beside one refusal is a worse reply than the refusal on its own.

1.106.031 July 2026

How many cm in a foot

how many cm in a footnothing30.48 cm
how many grams in a kilonothing1,000 grams
how many feet in a milenothing5,280 feet
how many seconds in a minutenothing60 seconds
The converter needs something to convert and this sentence names no amount. The amount is one, and saying "1 foot" out loud is what nobody does.
Not where the date grammar answers better: a month is 31 days in July and 28 in February, and turning that into a definition would be the wrong answer.

1.105.031 July 2026

Correlation, which needs two lists

correlation of 1 2 3 and 2 4 6nothing1
correlation of 1 2 3 and 6 4 2nothing-1
covariance of 1 2 3 and 2 4 6nothing2
Every reader in this section takes one list, which is right for every question it answers. The question people bring two lists to a calculator for is whether they move together.
Pearson's r, with the ends of its scale printed underneath, since r is on a fixed scale and the ends of that scale are the whole of how to read it.
A list that never changes has no spread to share, and nought would read as "no relationship" rather than "no question".

1.104.031 July 2026

Quartiles said in words

second quartile of 1 2 3 4nothing2.5
first quartile of 1 2 3 4nothing1.75
quartile 2 of 1 2 3 4nothing2.5
9th decile of 1 2 3 4 5 6 7 8 9 10nothing9.1
"q2" read and "second quartile" did not, and the words are what somebody who has not met the abbreviation types, which is exactly the reader who needs the answer.
The deciles come with them, being the same idea counted in tens, and they had no spelling here at all.

1.103.031 July 2026

The midrange

midrange of 1 2 3 4nothing2.5
midrange of 1 2 3 100nothing50.5
middle of 1 2 3 4nothing2.5
Halfway between the smallest and the largest, which is not the middle of the list and not the average of it: for 1 2 3 100 the median is 2.5, the mean is 26.5 and the midrange is 50.5.
All three were computable here and only two of them had names, though the midrange is what a thermostat and a weather forecast both mean by "average".

1.102.031 July 2026

How many numbers there are

count of 1 2 3 4nothing4
length of 1 2 3 4nothing4
count of 1 to 100nothing100
The one thing about a list this worked out for every other answer it gave and could not be asked for: every echo line here has ended in "of 7 numbers" since the list readers went in.
The count is also the one answer whose echo cannot name the count, since "count of 4 numbers" is the answer said twice. It shows the numbers instead.

1.101.031 July 2026

Solving for x

solve 2x + 3 = 11nothingx = 4
x^2 = 16nothingx = 4 or -4
x^2 + 2x - 8 = 0nothingx = 2 or -4
3(x+1) = 12nothingx = 3
A whole arithmetic engine, and no way to point it at the commonest sum anybody is ever set: the one with the answer missing from the middle instead of the end.
Solved by sampling rather than by rearranging: x is substituted with four numbers and the four answers say what shape the equation is, so this inherits every function and bracket the parser already knows.
Every answer is checked by putting it back in, and anything with more curve in it than a square is refused rather than answered wrongly.

1.100.031 July 2026

A hex literal where a number goes

what is 0xff plus 1nothing256
0xff * 2nothing510
0b1010 + 5nothing15
0b1010Sun, 29 Apr 202910
The bases grammar has read "0xff in decimal" since it went in. A literal is a spelling of a number rather than a conversion of one, so anywhere this takes a number it ought to take that spelling.
It did not, and the letters were read as a word: "0xff plus 1" went looking for a function called xff, and "0b1010" was read as an offset of a thousand and ten business days.

1.99.031 July 2026

Pi to a hundred places

pi to 10 placesnothing3.1415926536
e to 20 placesnothing2.71828182845904523536
phi to 10 placesnothing1.6180339887
tau to 10 placesnothing6.2831853072
A double holds fifteen significant figures and this page prints twelve, so "pi" here has always been 3.14159265359 and could not be anything else.
Asking for more had a wrong answer available: Math.PI padded out with the noise at the end of its binary expansion, which is the worst kind of answer to give.
So the digits are shipped, without apology. They are the one kind of constant that cannot go stale: the hundredth digit of pi was the same before anyone counted it.