Releases 1.60.0 – 1.98.0

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.

1.98.031 July 2026

The nth Fibonacci number

fibonacci 10nothing55 · 21 + 34 = 55
10th fibonacci numbernothing55
fib(10)nothing55
the 12th fibonaccinothing144
Counted from F(1) = 1, which is the numbering anybody saying "the 10th Fibonacci number" out loud means, and the one that makes the 10th 55.
Added rather than solved. Binet's formula gives a closed form and gives it wrong somewhere in the seventies, which is inside the range this holds, and seventy-eight additions is not slow.
The two it is made of come with it, since that is the definition and also the check.

1.97.031 July 2026

Is 28 a perfect number

is 28 a perfect numbernothingyes · 1 + 2 + 4 + 7 + 14 = 28
is 12 a perfect numbernothingno · 1 + 2 + 3 + 4 + 6 = 16, so it is abundant
is 12 abundantnothingyes
is 8 deficientnothingyes
The divisors have been listable since "factors of 360" went in, and this is what they add up to.
Perfect numbers are rare enough that a bare "no" is the answer almost every time, so the sum is always shown along with which side of perfect the number falls on.

1.96.031 July 2026

Is 100 a square number

is 100 a square numbernothingyes · 10² = 100
is 50 a square numbernothingno · between 7² = 49 and 8² = 64
is 27 a cubenothingyes · 3³ = 27
is 10 triangularnothingyes · 1+…+4 = 10
"is 100 prime" has been answerable since the primes went in, and this is the same sentence with a different shape of number in it.
A no comes with the two it falls between, which is the useful half of a no: "50 is not a square" says nothing that "50" did not.
A cube may be negative and the other two may not, so -8 is a cube and -4 is not a square.

1.95.031 July 2026

The square of 5, said that way round

square of 5nothing25
cube of 3nothing27
reciprocal of 4nothing0.25
inverse of 8nothing0.125
"5 squared" reads and "the square of 5" did not, though the second is how it is said when the number is the thing being talked about.
The root already read both orders, because "square root of 16" is the only order it has. The power read one and the reciprocal read neither.
The operand is a number, a bracket or a name and nothing looser: run to the end of the line, "square of 5 + 3" would come out as the square of eight.

1.94.031 July 2026

A logarithm with its base said out loud

log base 2 of 1024nothing10
log base 3 of 81nothing4
log 1024 base 2nothing10
log_2(1024)nothing10
log2 and log10 are the two bases nobody has to name, because they have their own spellings. Every other base had no way in at all.
The two named bases are still used where they are asked for rather than computed as a ratio of natural logs, since log(1024)/log(2) is 10 to within a rounding error and this page prints eleven decimal places.

1.93.031 July 2026

Rounding that says how much to keep

round 3.14159 to 2 placesnothing3.14
round(3.14159, 2)nothing3.14
round 1234 to 3 significant figuresnothing1,230
round 1234 to the nearest 100nothing1,200
round() has taken one argument since the function list went in, which rounds to a whole number: the one rounding nobody has to ask a calculator for.
Scaled, rounded and scaled back, so a negative count works as well as a positive one and is the same arithmetic rather than a second rule.
Significant figures and the nearest something are offered for "round" only, floor and ceiling to three significant figures being a sentence nobody says.

1.92.031 July 2026

The primes below a number

how many primes under 100nothing25 primes
prime numbers under 50nothing15 primes · 2 · 3 · 5 · 7 · …
primes between 100 and 200nothing21 primes
how many primes under 100000nothing9,592 primes
Every other prime question here is trial division, which is right for "is 5,381 prime" and wrong for this. So this is the one place a sieve earns its memory.
"under 100" stops short of it and "up to 100" does not, which matters for exactly the numbers that are themselves prime.
Listed where a list is readable and topped and tailed where it is not: nine thousand numbers in a row is not an answer, and neither is a count with nothing to check it against.

1.91.031 July 2026

A run of numbers written as its two ends

sum of 1 to 100nothing5,050
average of 1 to 100nothing50.5
stdev of 1 to 10nothing3.0276503541
sum of 1 to 1000000nothing500,000,500,000
The list readers take numbers one at a time, which is right for the six or seven anybody types out and useless for the shape this question is nearly always asked in. Nobody writes a hundred numbers to add them up.
Materialised where the run is small enough to be a list, so median, spread and percentile all get it for nothing.
Past that, only the readings with a closed form, since a million numbers held in memory to add in order is a slow way of doing arithmetic Gauss did at nine.

1.90.031 July 2026

How long the flight is

how long to fly from tokyo to londonnothing11h 8m
how long is the flight from new york to londonnothing6h 41m
flying time from paris to berlinnothing1h 28m
The distance has been answerable since the coordinates went in, and dividing it by a speed is the step everybody was doing afterwards.
900 km/h and half an hour on the ground, printed on every answer, because this is the one figure in the file that is deliberately approximate and it should not look measured.
No wind and no route: a real Tokyo to London flight is an hour or more longer westbound, the jet stream being worth several hundred kilometres an hour and nothing here modelling it.

1.89.031 July 2026

The antipode

antipode of londonnothing-51.51, 179.87 · 882 km from Chatham
antipode of madridnothing-40.40, 176.32 · 415 km from Auckland
antipodes of aucklandnothing36.87, -5.23 · 83 km from Gibraltar
The one place a distance question has a fixed answer for: every antipode is 20,015 km away, half a circumference.
The interesting half is where it lands, so the nearest place this file knows is named alongside.
Latitude flips sign and longitude turns halfway round, wrapped properly, or half the world comes back with a longitude past 180 that no map takes.

1.88.031 July 2026

The midpoint between two places

midpoint between tokyo and londonnothing69.02, 89.75 · 4,782 km from each
midpoint of new york and londonnothing52.37, -41.30
midpoint between paris and berlinnothing50.82, 7.64
Not the average of the two coordinates. Averaging Tokyo and London puts halfway at 43.6°N; the halfway point of the flight is at 69.0°N, because the route goes over the top.
The nearest place the file knows is named alongside, since a coordinate in the middle of Siberia says nothing on its own and "1,412 km from Tomsk" says where it is.
No new data: the coordinates are the ones the sunrise times already read.

1.87.031 July 2026

The earliest sunset

when is sunset earliest in londonnothingFri, 11 Dec 2026 · 3:52 pm
latest sunset in londonnothingTue, 23 Jun 2026 · 9:23 pm
earliest sunrise in londonnothingWed, 17 Jun 2026 · 4:43 am
when is sunset earliest in tokyonothingSun, 29 Nov 2026 · 4:29 pm
The one everybody gets wrong. Sunset is earliest a fortnight before the solstice and sunrise is latest a fortnight after it, because solar noon itself drifts later through December.
In London the earliest sunset is on the 11th and the shortest day is the 22nd, and the answer names both, since those eleven days are the whole point of asking.
Compared by the clock on the wall, which is what "earliest" means, and costs an offset lookup a day: the same comparison done in UTC would be a comparison of something else.

1.86.031 July 2026

The longest day and the shortest

longest day in londonnothingSun, 21 Jun 2026 · 16h 38m
shortest day in londonnothingTue, 22 Dec 2026 · 7h 49m
longest day in sydneynothingTue, 22 Dec 2026 · 14h 25m
longest day in longyearbyennothing127 days of 24h 0m
Everything this needs has been here since the sunrise equation went in, and the question could only be answered by asking three hundred and sixty-five others.
Not quite the solstice: that is one instant for a whole hemisphere, and the longest day somewhere is whichever of that place's days holds the most daylight.
Inside the circles there is no one longest day, so the stretch is given instead: naming the first of 127 identical days would be picking one arbitrarily and printing it as the answer.

1.85.031 July 2026

The sun question with the question left in

how much daylight is there in london todaynothing15h 50m
how long is the day in londonnothing15h 50m
what time is sunrise in londonis sunrise isn't a place I know5:13 am in London
what time does the sun set in tokyodoes the sun set isn't a place I know6:53 pm in Tokyo
The keyword had to open the line or close it, and in real English it sits in the middle with a question on either side.
The keyword may be written apart and may be a verb: "what time does the sun rise" is the same question as "sunrise" and had none of its spelling.
The clock grammar reads the same opening words and is not this question, so it hands over rather than blaming a place nobody typed.

1.84.031 July 2026

Sunrise tomorrow

sunrise tomorrowtomorrow isn't a place I know5:47 am
sunset tomorrowtomorrow isn't a place I know8:20 pm
sunrise on fridayon friday isn't a place I knowFri, 24 Jul 2026
sunrise in junejune isn't a place I knowMon, 1 Jun 2026
A day and no place is how the question is asked from wherever you are standing, and every word was being offered to the place resolver instead.
"tomorrow isn't a place I know" is true and unhelpful, and it blames the day for not being a city.
A place still wins where there is one, and the day reads on exactly the terms it already read on after a place.

1.83.031 July 2026

A clock question with a length in it

what time will it be in 3 hours in tokyowill it be in 3 hours isn't a place I know12:00 pm in Tokyo
what time will it be in tokyowill it be isn't a place I know9:00 am in Tokyo
what time was it 3 hours ago in londonnothing10:00 pm in London
The clock grammar has moved a time by a span since it learned "5 hours after now", and only ever from a clock reading.
Asked the way it is spoken, with a question then a length then a place, nothing matched and the place resolver was handed the whole tail and blamed it for not being a city.
The tense may be any of the three, since the question gets asked in all of them.

1.82.031 July 2026

Is anybody at work there

is it business hours in tokyonothingYes, 9:00 am Fri in Tokyo
is it business hours in londonnothingNo, too early, they open at 9:00 am
is it working hours in new yorknothingNo, too late, they shut at 5:00 pm
is the office open in tokyonothingYes, 9:00 am Fri in Tokyo
Every part of this was here and none of them were joined up: the clock says what time it is there, the weekday rule whether it is a working day, the holiday tables whether the country has the day off.
Nine to five, Monday to Friday, said out loud on every answer. It is a convention rather than a fact, it is wrong for a Spanish afternoon and wrong for a trading floor, and a number that arbitrary has no business being implied.

1.81.031 July 2026

When the clocks change

when do the clocks change in londonnothingSun, 25 Oct 2026 · 2:00 am → 1:00 am
when does dst start in new yorknothingSun, 14 Mar 2027 · clocks go forward
when do the clocks change in sydneynothingSun, 4 Oct 2026 · clocks go forward
clocks change in tokyonothingTokyo doesn't change its clocks
Found by looking rather than by rule: the runtime carries the whole tz database, so a year is walked a day at a time and the gap halved wherever two neighbouring days disagree.
A table of rules written here would be a second opinion on data the file already has, and the second opinion is the one that goes out of date.
The direction is honoured where it is named, since "when does dst end" is a question about the autumn and the spring is a different answer.

1.80.031 July 2026

Which time zone a place is in

what timezone is boston innothingEastern Daylight Time · America/New_York · UTC-4
what time zone is kolkata innothingIndia Standard Time · UTC+5:30
timezone of sydneynothingAustralian Eastern Standard Time
A place has resolved to a zone since the first clock answer and the file never once said which zone, except as an id on an echo line nobody reads it off.
Somebody filling in a calendar invitation wants "Eastern Time", not "America/New_York" and not "UTC-4", though both are worth having underneath.
The names come from the runtime, which carries them for every zone it knows. Nothing is shipped, which is the bargain the calendars and the coordinates already take.

1.79.031 July 2026

The offset of one place

utc offset of tokyonothingUTC+9
utc offset of kolkatanothingUTC+5:30
how many hours ahead is tokyonothingTokyo is 13 hours ahead of New York
how far behind london is new yorknothingNew York is 5 hours behind London
Both numbers were on the page already and neither could be asked for: the offset is on the echo line of every clock answer, the gap is the headline of every two-place one.
A question about one place had to be padded out with a second before this would take it, and the second one people mean is where they are sitting.
The word asked with is answered even when the sign disagrees with it: asking how far ahead Los Angeles is deserves being told that it is behind.

1.78.031 July 2026

Holidays that fall on a weekend

how many holidays fall on a weekend in 2027 in the usnothing3 of 11 national holidays
how many bank holidays fall on a weekend in 2027 in the uknothing2 of 8 bank holidays
how many holidays fall on a weekend in the usnothing1 of 11 national holidays
The one thing a list of next year's holidays does not give you by being read out.
The count comes with what is done about it, since each country's answer is already here: the States move the day to the nearest weekday, Britain and Canada add one after.
A holiday on a Saturday costs a day in the States and costs nothing in Britain, and a bare count does not say which.

1.77.031 July 2026

Holidays in a month

holidays in july in the usnothing1 national holiday · 4 Jul Independence Day
holidays in the us in julynothing1 national holiday
holidays in 2027 in the usnothing11 national holidays
holidays in august in the usnothingno national holidays
A year could only ever be tacked on the end, so the two halves of the question had one order between them and the month had no place at all.
Whichever half names a country is the country and the other is the stretch, so both orders read.
A month with nothing in it says so in words. "0 national holidays" is arithmetic where a sentence belongs, and silence reads as not understood.

1.76.031 July 2026

Is today a holiday

is today a holiday in the usnothingNo · a working day in the US
is 4 jul 2026 a holiday in the usnothingYes, Independence Day
is 28 dec 2026 a holiday in the uknothingYes, Boxing Day
is christmas a holiday in the uknothingYes, Christmas Day
"what day is 4 jul 2026 in the us" answered it sideways, by naming the weekday first and mentioning the holiday after. Somebody deciding whether to send an invoice wants the word yes.
The holiday a day is, against the day an office takes off for it: the 4th of July 2026 is a Saturday and the States shut on the Friday, and the line says both.
A holiday belongs to a country, so being asked without one is a question with a hole in it rather than a question about nowhere.

1.75.031 July 2026

The next holiday

next holiday in the usnothingLabor Day · Mon, 7 Sep 2026
next bank holiday in the uknothingthe summer bank holiday · Mon, 31 Aug 2026
last holiday in the usnothingIndependence Day · Sat, 4 Jul 2026
next holiday in tokyonothingMountain Day · Tue, 11 Aug 2026
The list of a whole year was answerable and a whole year of dates is not what somebody wondering when the next long weekend is has asked for.
Strictly past today. A holiday you are standing in is not the next one and is not the last one either.
Where the day off is not the day itself, both are shown: one is when it is, the other is when the office shuts.

1.74.031 July 2026

A week on Tuesday

a week on tuesdaynothingTue, 4 Aug 2026
tuesday weeknothingTue, 4 Aug 2026
a week todaynothingFri, 31 Jul 2026
a fortnight on tuesdaynothingTue, 11 Aug 2026
How a fortnight's notice is given in half the English-speaking world, and a shape this had no reading for at all.
Read whole, since a word at a time "tuesday week" is a Tuesday with a stray unit behind it and "a week on tuesday" is a length with a day it cannot use.
Said on a Friday, "friday week" is seven days out and not fourteen: the day it names is the coming one and today counts as it.

1.73.031 July 2026

The Sunday before Christmas

the sunday before christmasnothingSun, 20 Dec 2026
the friday after easternothingFri, 2 Apr 2027
the monday before new yearnothingMon, 28 Dec 2026
An ordinal that is always "first" is one nobody says out loud, and requiring it was the whole difference between the sentence people write and the sentence this could read.
Only in front of a step away from a day: "sunday of june" names four Sundays and is not this question, so it stays what it was.
The article was not being skipped on the way in either, so "the 3rd friday of june" now reads as well as the version without it.

1.72.031 July 2026

A day counted from a named day

2nd sunday after easternothingSun, 11 Apr 2027
3rd friday before christmasnothingFri, 4 Dec 2026
1st monday after new yearnothingMon, 4 Jan 2027
3rd business day after christmas in the usnothingWed, 30 Dec 2026
A month was the only stretch this count could walk through, so every question anchored on a moveable feast had no shape at all, Easter being the reason half a calendar moves in the first place.
Strictly away from the day: Easter is a Sunday, and a first Sunday after it that answered Easter would be a week wrong every year.
"last friday before christmas" is the first one going backwards. "last friday after christmas" names no day at all and is left to say nothing.

1.71.031 July 2026

How far into a period we are

how many days into the year are wenothingDay 205 of 365
what day of the year is itnothingDay 205 of 365
how far into the quarter are wenothingDay 24 of 92
how many weeks into the year are wenothingWeek 30 of 53
"day 200 of 2026" has been readable from the other direction since the ordinal day went in: it answers which date that is. This asks which ordinal today is.
Weeks are counted in sevens from the first of the period and say so, since that is not the ISO week the page prints elsewhere.
Months into a month is not a question, and answering "1 of 1" would be a way of pretending it was.

1.70.031 July 2026

Since points backwards

days since christmasnothing211 days ago
days since easternothing110 days ago
days since halloweennothing266 days ago
days since mondaynothing4 days ago
Every recurring name in this file reads forwards, and "since" was filler, so it was thrown away and Christmas resolved to the coming one: a date nothing can have happened since.
This grammar already has a word for the last one, so the word is spelled "last" here and read by the rule that has always read it.
A date is not an occurrence of anything, so "days since 1 jan 2020" falls through to the filler the word has always been.

1.69.031 July 2026

What is left of a period

days left in julynothing7 days
how many days left this yearnothing160 days
working days left this year in the usnothing109 working days
how many mondays left this yearnothing23 Mondays
weeks left in the yearnothing22 weeks 6 days
"days in july" answers 31 on the 31st. It is true, and it is not what anybody standing inside July is asking.
Counted from tomorrow, because a day you are standing in is not a day you have left.
A period that has not started keeps all of itself, so "days left in september" is 30. One that has ended says so, which beats answering nought to a question about May asked in July.

1.68.031 July 2026

A counted working day is a day the office is open

3rd business day of july 2026 in the usFri, 3 Jul 2026Mon, 6 Jul 2026
1st business day of january 2027 in the usFri, 1 Jan 2027Mon, 4 Jan 2027
3rd business day of july 2026Fri, 3 Jul 2026Fri, 3 Jul 2026
One question asked two ways gave two answers. "1jul2026 + 3 business days in the us" stepped over the Friday the States take for the 4th; "3rd business day of july 2026 in the us" landed on it.
Without a country it is weekends and nothing else, exactly as before, because without one there is no telling whose holidays to skip.

1.67.031 July 2026

The month next door

first monday of next monthnothingMon, 3 Aug 2026
last day of next monthnothingMon, 31 Aug 2026
last day of last monthnothingTue, 30 Jun 2026
first day of next yearnothingFri, 1 Jan 2027
Nobody plans the first Monday of the month they are already most of the way through. The month a rota or a rent day is about is the next one, and it was the one this could not be asked about.
Read as a stretch rather than left to the offset grammar, which answers "next month" with this day next month and leaves no month to count inside.

1.66.031 July 2026

The month said the way it is said out loud

last day of this monthnothingFri, 31 Jul 2026
first day of this monthnothingWed, 1 Jul 2026
last day of this yearnothingThu, 31 Dec 2026
3rd friday of this monthnothingFri, 17 Jul 2026
"last day of the month" answered. "last day of this month" did not, because the article was the only word being skipped over on the way to the period.
It carries no date of its own, so what follows it is read exactly as "the month" and "the year" already were.

1.65.031 July 2026

The period we are in

what week is itnothingWeek 30 of 2026
what quarter is itnothingQ3 2026
what month is itnothingJuly 2026
what year is itnothing2026
Every period this reads answered the opposite question. "week 30" says which Monday week 30 opens on, "q3" which day the quarter starts, "july" the first of the month, and none of them could be asked which one we are standing in.
The span comes with it: 20 Jul 2026 to 26 Jul 2026, day 5 of 7, 2 days left.
The tail of a longer question stays with whoever owns it, so "what year is it in the hebrew calendar" is not answered 2026.

1.64.031 July 2026

An ordinal where only the first was allowed

10th prime under 300nothing241
3rd prime after 100nothing107
100th primenothing541
3rd largest of 5 3 9 1 7nothing5
2nd smallest divisor of 360nothing2
2nd largest prime factor of 360nothing3
Four grammars had the first and the last and nothing in between. The walk that finds the next prime was only ever asked to take one step.
Bounded by the work, not the count: the 100,000th prime is 42ms because the numbers tested are small, ten primes above a trillion is 141ms because each test has six digits' worth of factors to try.

1.63.131 July 2026

The footer stays where you can see it

The version and the other pages were three screens down this page. The footer is held at the foot of the window now, on the prose pages.
It still takes its own space, so the oldest release is not printed under it. The calculator keeps its own: the suggestion list opens downward into the same pixels.

1.63.031 July 2026

How long from one sun time to another

sunset to dawn in reykjavik 21dec2026nothing18h 34m · 3:30 pm → 10:04 am the next day
sunrise to sunset in londonnothing15h 50m · 5:13 am → 9:02 pm
29 mpg in litres per 100 kmnothing8.11085 l/100km
500 km at 8 litres per 100 kmnothing40 l
Both fuel spellings were in the table all along. The reader could not see a unit with a space or a digit in the middle of it.

1.62.031 July 2026

Ten countries outside Europe know their holidays

holidays in japannothing14 national holidays
what day is it in mumbaiFriday in Mumbai · a weekdayFriday in Mumbai · a working day in India
what day is 6 may 2026 in japanWednesday · a weekdayWednesday · Children's Day, observed in Japan
India, Mexico, Russia, Japan, Australia, South Africa, New Zealand, Kenya, Ghana and Kazakhstan. Thirty-two countries.
757 rules checked against the published lists, none drifted. The check caught two of mine: Ghana's Farmers' Day is the first Friday of December, not the 4th, and Mexico moves a weekend holiday to the nearest weekday.
China, South Korea and Vietnam are out. Their year hangs off Lunar New Year, which is a new moon rather than a rule. Japan is here without its two equinox days, which the observatory announces. India has the three days fixed by statute.

1.61.031 July 2026

The holidays a country keeps, listed

how many national holidays are in canadanothing9 national holidays
bank holidays in the uknothing8 bank holidays · 1 Jan New Year's Day · 3 Apr Good Friday · …
holidays in torontonothing9 national holidays
No new data. These are the rules the working-day count already skipped, read out.
Counted in the noun you asked in: 8 bank holidays, 11 federal ones.

1.60.031 July 2026

The difference between two clocks, asked for outright

time difference between jerusalem and parisnothingParis is 1 hour behind Jerusalem
what is the time difference between tokyo and londonnothingLondon is 8 hours behind Tokyo
time difference between dc and nycnothingno difference, they keep the same clock
tokyo to london still answers with the clock and puts the gap underneath. Asked this way the two swap lines.