Three Words – Any Place

Here’s an amazing new tool. what3words has identified every one of the 57 trillion 3mx3m squares on the entire planet with just three, easy to remember, words. My office, for example, not my building but my office, is token.oyster.whispering. Tyler’s office just down the hall is barons.huts.sneaky. (Especially easy to remember if you recall this is Tyrone’s office as well.)

Every location on the earth now has a fixed, easily-accessible and memorable address. Unpopulated places have addresses for the first time ever, of course, but now so do heavily populated places like favelas in Brazil where there are no roads or numbered houses. In principle, addressing could be done with latitude and longitude but that’s like trying to direct people to web sites with IP addresses–not good for humans.

Algorithms have assigned words to avoid homophones (sale & sail) and to place similar combos far from one another to aid in error detection. Simpler, more common words are used to address more populated areas and longer words are used in unpopulated areas.

Moreover the three word addresses are available not just in English but in French, Spanish, Portuguese, Swahili, Russian, German, Turkish and Swedish with more languages on the way. The addresses in other languages are not translations but unique 3 word addresses in those languages.

All of this is available in a small app so that it can be used even offline on a simple smartphone. Find your address here.

Hat tip: The Browser.


Definitely needed in Japan where they don`t even have addresses.

Japan has addresses. They're just not linearly ordered the way street numbers are. With GPS, which you need to use this tool as well, it's not hard to find the location of a given address.

I heard that they have strret numbers, but they are assigned in order of when the building was constructed. I seem to recall Korea also adopted this scheme during the Japanese occupation.

I wonder if this system elevates the status of having a new house - could be a partial explanation for why there is so much turnover in Japanese housing.

They are not street numbers, they are block numbers. Streets are unnamed, blocks have names. The first house in a block is named "block-name 1" and counting up.

Japan has a "hierarchical address system".

It starts with city (such as "Tokyo"), then ward (such as "Chuo"), then subarea (such as "Ginza"), then numbers for further subarea, block number, and house number (where house number is typically in order of construction).

You state these in reverse order, so:

5 (further subarea) 2 (block) 1 (house) Ginza (subarea) Chuo (ward) Tokyo (City)

How about the height dimension? Tyrone's office might be on top of Tyler's.

The web site suggests using something like: 6th Floor, barons.huts.sneaky.

They need to edit that page right now to include links to the App Stores. They sold me and then had no call to action!

It seems to be a proprietary system, which means that it will never be a good standard. Both because it will be inconvenient - you can't always access their API for various reasons - and because it doesn't really make any sense to place something as important as global positioning on the whims of single corporation that can decide to do whatever they want with the system.

Probably the idea is to sell out to Google.

Google already has a competitor:

Google's is better, because it works with traditional addresses hierarchically.

I was confused about how they got 55 trillion combinations, when they said they started with a list of 25,000 words and threw some out. Turns out they use declensions and conjugations, not just the base words.

Unless you live or work in a building with more than one floor. Since it identifies squares of land area, everyone shares an address with everyone above or below them.

Use for words to uniquely identify every 3x3x3 cube of space

The web site suggests using something like: 6th Floor, barons.huts.sneaky.

Which increases the needed words exponentially, and is pointless when very few occupied places on the planet are more than 3 meters above the surface.

Doesn't it just require a fourth word position? Even "nth floor" can be considered a word.

AFAIK, no requirement that each word can only be in one position.

That said, I agree it isn't (yet) that useful, except perhaps in large urban areas (e.g. Hong Kong) that probably have a proper address system already.

foot.foot.foot is north of Anchorage, Alaska.

I think it's actually orders of magnitude most useful for densely populated urban areas that don't have an address system. It's secondarily useful for places with a confusing address system (repetitive, similar or nearly identical street names, for example.)

They should have distributed the short and common words in the most densely populated areas. So foot.foot.foot should be in New York or London, and rutabaga.rutabaga.rutabage should be in Alaska.

New York and London are far from the most densely populated places. Foot.foot.foot would be wasted on them.

Worst use of "exponentially" ever. The increase is logarithmic, literally the opposite of exponential.

Just looked at the map. Funny, but it makes the mailman crazy. The house next to mine has a completely different 3 word address Vs "X street 101" for my house and next house "X street 103".

This is an interesting topic. We don't find a thing because we simply know it's position. We find it because we know: a) where we are (785 X street, rail station), b) where the thing is (224 Y street), c) the position of the thing relative to the other things (3 blocks from the river). This system makes part "b" really simple, but parts "a" and "b" of the orientation process are not understandable to humans.

However, if this is aimed at robot cars or delivery drones it's really good. You input your 3 word home address to the robot taxi and you'll get home even if you don't understand how the system works.

Wouldnt robots be better off using GPS coords?

Maybe very marginally. GPS coordinates, or an small intervening app...virtually no difference to a computer.

USPS will be the last adopter. Who cares?

...good point. Humans need a geographical identification system that easily "relates" one position to another. This "amazing" new 3word system falls well short.

Note that there are many existing coordinate systems beside Lat/Long.
Universal Transverse Mercator (UTM) is pretty good, commonly used worldwide, and easily beats the overall utility of this 3word system.

By robots, for robots. Behold the wondrous future of technology.

Lat-long is much better for robots. Location x is n miles NNE from location y, Location y has coordinates of AA°BB′CC.DD″N EE°FF′GG.HH″W can be calculated just with the information given as opposed to having to reference a lookup table.

From wikipedia:

The CN Tower is at 43°38′33.24″N 79°23′13.7″W, which is in UTM zone 17, and the grid position is 630084m east, 4833438m north. Two points in Zone 17 have these coordinates, one in the northern hemisphere and one in the south; one of two conventions is used to say which:

Oh yeah, that's much easier to remember than select.threaten.shelters.

select.threaten.shelters is easy to remember, but otherwise entirely worthless. I know that 43°x79° is a little bit north and a very long ways west of me.

I don't know, I prefer "The CN Tower."

Until the naming rights change, and all of your references are out-of-date...

Can't that be solved by just sorting the words and using them in order?

worda.wordb.wordc is next to worda.wordb.wordd and so on.

Could get more fancy by varying the 3rd position east-west and the 2nd position north-south, but my guess is the designers already thought of that and found some reason to dismiss it (er, grid coordinates on a sphere).

They deliberately diddn't do that, to help correct errors with transpositions.

If is in one place then is very far away. Under the assumption that you usually want to look for places reasonably close to you (on the same continent), this lets them detect some errors because the resulting place is very far away, enabling them to suggest some addresses which are similar to the one you typed but closer to you, letting them ask, "Did you mean ...?".

If you use the "2954 North Weston Road" way, then if you typo and get "2594 North Weston Road", you have no (easy) way of finding that mistake.

Precisely. Manhattan's use of numbered streets and avenues on a grid is one of the things that makes it much easier to navigate than say Boston. But even Boston is a paragon of orderliness compared to these randomized words.

There's yet another vital piece of geographic information the these 3words utterly fail to convey: where a location is within the hierarchical scale of locations. The Chinese are better than this than the US is; they locate themselves using a scale that works like this: world, China, Beijing, abc Street, xyz number. Whereas in the US we put the state in between the city and zip code rather than putting the information in order.

Humans don't start out trying to find Carrow Hall 2, or some other 3x3 square. And if you're told "oh that's barons.hut.sneaky", what good does that information do you? (Unless you've got their app, are connected to the web, etc.)

As others have mentioned, if I want to know where Tyler's office is, barons.hut.sneaky is a lot less useful than knowing it's at GMU, or in Carrow Hall, etc.

Even places with chaotic addresses such as Tokyo will find this of limited usefulness. If I want to go to the Ginza district, then I want to go to the Ginza district and not figure out which of the hundreds of thousands of 3x3 squares is the one that I want to go to.

Interesting but useless at the same time.

negatives: Apparently, the words are chosen quite randomly. This means that one cannot know whether two locations are closer or further apart just based on their 3word codes. And if Lat Long coordinates are not good for humans, this isn't any better either. Because these words are chosen randomly, people don't relate to them and will probably not use it.

positive: Perhaps this approach does bring some advantages if this is more efficient way for computers to store address information and where there is no need for a human interface.

You seem to be missing the point of the idea quite a bit. The three words are easier for people to use. Similarly how often does one judge the distance between two addresses in a city solely from the addresses without knowing where they are in advance?

The advantage seems to be, if the scheme is added to Garmins and TomToms, that you can give the three words to your cabby in a city without proper addresses and he'll be able to route directly to the location you want to go to.

I routinely use address numbers to judge relative distances. 100 Main St. is much further away from 10000 Main St than it is from 1000 Main St.

Fantastic, but I don't understand why the words are attributed randomly, rather than nested geographically.

Its probably for the same reason that the characters in "reservoir dogs" werent allowed to chose their own code names: "cause everyone chooses Mr. Black and nothing ever gets done."

I think Stuart meant that it would make sense for the two offices in the same hall to share the first two words, and not the last word. Then one could see at a glance that token.oyster.whispering is near to token.oyster.sneaky .

They probably did this intentionally, to make them easier to remember; if I'm telling you to go to an office, if I only remember one of the words ("I think it had a baron in it?") there's probably only one location that fits nearby ("oh, you mean barons.huts.sneaky.").

Telephone area codes were delibrately chosen to be widely geographically distinct to avoid errors caused by confusing them, so 212 is New York and 213 is Los Angeles.

That was more about the rotary dial phone. Cities wanted the quickest 3 number combos to dial which is how NY got about the best one.

Right. They were allocated such that the largest populations (and thus presumably the most frequently dialed areas) had the smallest dialing time on a rotary phone. Hence 212 for NY, 213 for LA, 312 for Chicago, and so on.

This seems to miss the point, because telephone numbers are nested just as people propose for the words: area code - local exchance - number.

Having similar values on the same level not be adjacent makes sense, though.

Nesting would require more words.

No it wouldn't; it could be accomplished simply by rearranging the existing combinations in a different configuration.

If one wanted the nested area boundaries to also correspond to geographical or political boundaries, it would require slightly more words. But that isn't necessary for the idea to be useful.

Pitcher's mound at Yankee stadium - spray.slimy.closes

Can you imagine telling a friend,

"I'll meet you at

slimey.stinking.oyster for a bite

and we can later go to

barfing.farting.poop for a beer."

Buckingham Palace: mole.plot.voted

White House (south lawn): employ.remove.files

The Kremlin: crafted.appendix.voltage


It is of course an encoding of GPS coordinates for human communication. It is clever as that. But it really presumes spoken use. In this day and age we need an easy and interoperable geotag. It can be totally opaque to the user. It can just look like a ?. As long as it can be sent from phone to tablet to computer to car to emergency services we are good.

As we type these comments we should have a location tag button available.

This is one of those areas where we let Apple and Google fight though, rather than demanding a solution.

Is there a reverse directory? Who's got bonkers.waste.time?

Yes there is, that address isn't found however. Give it a try.

silly.weeks.wasted is ENE of Butte, Montana

It is accurate. If you can ride a bike up the trail that starts at deck.compounds.meal you're in good shape.

I think it would be useful to say "I'm at smelly.plate.anaconda", instead of "I'm in the exhibit hall, over by Cardhalla." Or "Navigate to seven.spoon.scissors" instead of going to "Warner Park" and then trying to find the ice pond. Or, where is my phone? Oh, it's at dynamic.paper.eskimo. Found it. Or instead of Target's app telling you what aisle something is in, it could say "It's at post.seldom.entirely". This seems like a great idea.

Right. Step one, commit all possible three word combinations to memory. Step two, profit!

There is no memory involved in any of these examples, except the memory it takes to transfer info from a phone call or app into what3words. Perhaps you don't have a smartphone.

I guess I have been assuming that the app will be able to navigate you to the right spot with a compass or floor map or something.

ok, replace that with "step one, outsource all your mental processing to a smartphone." Looking forward to a bright future of several decades operating as an iphone's tamagatchi!

Glympse looks like a good enough way to share current location, but the don't see a way to map-select a future place with what3words accuracy.

Combined, the two would be a killer app

I see that glympse lets you paste in decimal latitude and longitude, so a two-app solution works for meet me "there."

If you are copying and pasting anyway, decimals are fine.

Since it's based on 3mx3m grids, a pretty standard american single-family house in suburbs could have a lot of unique ids associated with it. I looked up my address, zoomed way in, and you can see a wide variety of choices.

If it catches on, they could make a fortune selling places like general.motors.home. Right now it's in northern Quebec, miles from anything at all. Reminds me of the early days of internet domain names.

Quite right. We have a modest split level house and there are 27 unique identifiers for it. I'm not sure what the value is to having so many identifiers since those only deal with the two dimensional footprint of the house. I could have the same 2-D foot print but have several stories leading to greater living square footage. this is a curiosity at best.

That's right.

You get to choose whether to pick the one you can remember best, or the one representing your front door (which is presumably where you'd want people to be routed to, rather than "the centre of your house" or whatever).

Sometimes it's annoyingly precise, but at other times (e.g. identifying specific exits of a subway station or a stadium), it could come in handy, I suppose.

TIL: My house does not exist on these maps, but it was built in the 1940's.

Not sure why this is a good idea. Numbers severely dominate words when it comes to data processing. For humans this is of very limited use - almost no use, because the address itself conveys zero information - essentially substituting words for GPS coordinates which is charming but am struggling to see why it is better than what exists.

Number do not dominate when it comes to transfer via human means. I think that is what this is for. When I tell people to meet me somewhere, I usually drop a pin in a message. This would allow me to say three words (if I can't drop a pin) and they could navigate to that instead. I see the advantage.

Since this is only in 3m x 3m squares, I can imagine people searching their entire property for the best three words to use. It's probably better to say you live at than squishy.mole.hole.

If you're planning to sell, be advised that location.location.location is already taken-- it's an apparently undeveloped tract near the unpromisingly named Lake Sukkozero in Russian Karelia.

Along similar lines, is in western Alaska.

And ... does not exist.

Yup, I'm definitely using haunted.archduke.galloped over doodle.objections.noses if this catches on.

They should make an open source version available, so 8th graders can replace the lexicon with dirty words. Hours of entertainment.

Note that "either.lake.promising" (10 Downing street, London) and "lake.either.promising" (nowhere in particular outside of Addis Ababa) are both used. So be careful when targeting your missile.

I wish they would stop trying to program me and get back to work on the computers.

Could there be a market for these? For example, if thankful.cobras.smooching is on my property, could a cobra breeder pay the company (or me) to swap it with one of his squares?

And people said peak stupidity would never be reached!

I don't think it will, Bob. We can be stupider than this.

At least it's keeping the academics amused. They have a lot of time to kill.

To summarise, the purpose of the app is to precisely define a discrete 3 x 3m location in a form easily digestible to humans. It then interacts with a gps based mapping service to find that precise location. The process of defining the location and navigating to it are deliberately, completely divorced from each other - that is the essence of the innovation. Does this serve a commercial purpose? Time will tell. The most promising applications would seem to involve human courier or service deliveries to smart phone enabled targets that are hard to define by street address or fixed for only short time periods. Seems like a reasonable low probability, high potential return investment case given the obvious network properties if commercially adopted.

I'm torn between enthusiasm and concern since I've read both James Scott and Hernando de Soto.

Number to non-number Ramanujization of information proceeds apace. While the days are gone when the desert fathers could recognize and be gratefully recognized by nearly every local scavenger bird or mammal for fifty miles around, there have been, not so long ago, wonderworkers who distinguish every individual bee that flies and crawls to and from the hive at the edge of the beloved garden or field of clover. If the rest of us loved numbers as we loved words this impressive "app" would not be necessary. From the city of bread to the city of peace it was, maybe (from the closest northern point in the prior city) possibly eighty three steps up, four level, 584 down, 7 level on a dry day, then in the middle of the pathway three times 585 up and 17 levelish (assuming no ruts from recent rain), then down again and so on. If we loved numbers as Ramanujan did we could do without this application's needless but charming use of words. As even Housman admitted, Manilius versified numbers better than he, one of the four best Latinists of his day, could ever have done: and Pushkin (even now ungoogleable on this site - who'd have guessed, years ago?) from time to time threw a multi-syllable number or two into previously unknown heights of laddered verse as well.

Use the familiar to replace the unfamiliar. The best of us (not me, not by a long shot) did this in a different way between one or two thousand (desert fathers) years ago or a few decades (Solanus Casey) ago. The "math gene" could be more developed than it is. Last time I sat at the DMV I tried to slowly imagine the walk from Bethlehem to Jerusalem. Ramanujan, if we were lucky, would have written valuable essays on informatics. While considered a mere mathematician, Kolmogorov's knowledge of old Russian poetry was as intense as the classical specialist Housman's was of old Latin poetry; as a mere hobbyist, he was fortunate to live only a few generations after, and to speak almost the same language, as a very very well-spoken poet (and one who disavowed his worst lines, an example many of us should follow). Previous comment should have "humble wonderworkers" instead of "wonderworkers" and "well-laddered verse" instead of "laddered verse."

Use the familiar to replace the unfamiliar. The best of us (not me, not by a long shot) did this in a different way between one or two thousand (desert fathers) years ago or a few decades (Solanus Casey) ago. The "math gene" could be more developed than it is. Last time I sat at the DMV I tried to slowly imagine the walk from Bethlehem to Jerusalem. Ramanujan, if we were lucky, would have written valuable essays on informatics. While considered a mere mathematician, Kolmogorov's knowledge of old Russian poetry was as intense as the classical specialist Housman's was of old Latin poetry; as a mere hobbyist, he was fortunate to live only a few generations after, and to speak almost the same language, as a very very well-spoken poet (and one who disavowed his worst lines, an example many of us should follow). Previous comment should have "humble wonderworkers" instead of "wonderworkers" and "well-laddered verse" instead of "laddered verse."

Instead of using an address for a generally wider area, why would anyone uses an address that can be used by a cruise missile??

This isn't anywhere near as neat as my giving the International Star Registry $54 and having them name one of the billions of stars after my girlfriend. Space travelers millennia into the future will be referring to her when they point their rocket at that pinpoint of light on the other side of the universe.

I am inspired in the hope you have, both in faster.than.light speed and in the hope that the Chinese astronauts will use a list they can't pronounce.

Keep hope alive, Chuck. Don't listen to them when they cite unfounded rumors. Rumors like: American astronomers ignore the registry during the search for earth-like planets.

They could achieve even more compression if they used Hexadecimal numbers instead of names... geeks would love it more too.

If only there were a way to identify any location on earth by numbers...

The clear use for me is drone delivery. I've wondered how Amazon could deliver to a street address autonomously without landing in a pool or parking space, or injuring some poor dog. But if I can specify a 3m x 3m square for a drone to deposit my package, that might work. Place my order... and deliver to fuzzy.squash.dimples.

Comments for this post are closed