Bram Cohen is not only the creator of BitTorrent, the protocol responsible for between 25% and 55% of all internet traffic, but also the creator and co-creator of a number of great puzzles that he sells on his Shapeways Shop. Intrigued by yet another Shapeways Community Member with their own wikipedia article (5 and counting) and the obvious dichotomy between someone that has done more than almost anyone else to make Intellectual Property immediately and widely available for free while whilst now selling his own IP online, I decided to interview Bram.
Joris: Bit Torrent makes it possible for anyone to download content from the web. People critical of it say that it is a tool for infringing Intellectual Property. At the same time you have a Shapeways Shop where you basically sell models based on your own intellectual property? Do you see the duality?
Bram: I have a lot less interesting in the subject of intellectual property
than most people seem to expect. I have an interest in networking
protocols, and also one in puzzles, both of which happen to bring up
intellectual property issues, but I deal with such issues of necessity,
not because I particularly care about them. There is the interesting
question of how to get puzzles produced, and also how to try to make
money off of their production, which ideally I’d like to do, but that’s
a secondary issue, since obviously it’s impossible to make money off a
puzzle which people aren’t very interested in even if it’s free.
I know you’re not really interested in the intellectual property issue but…How would you feel if there was a “The Shapeways Bay” that offered your designs as a free download?
In some sense that already exists and I post my own stuff to it – the
‘Puzzle will be played’ burr site has a huge collection of burr
puzzles, and most of my experience with burr puzzles comes from
rebuilding most of the puzzles from that site in Burrtools. Only a tiny
fraction of all puzzles have any commercial value whatsoever, so piracy
isn’t really an issue for them. Among the puzzles with a little bit of
commercial value, the few people and companies who make them are
generally quite strict about only producing things with permission,
because the community is small enough that there’s little gain and a
lot of potential ostracism from doing otherwise. For the rare puzzle
which has so much commercial potential that it might attract real
knockoffs, the two approaches are to either make a brand-based premium
version, as the Rubik’s Cube does these days, or to ramp up supply
ahead of the knockoffs while the fad moves along, as the Rubik’s Cube
completely failed to do when it was first introduced. Patents don’t in
practice help all that much. All of that is in the ‘good problem to
have’ category though – most mechanical puzzles simply fail
commercially, and it’s unusual for one to succeed enough for knockoffs
to be a concern.
So how did you come to make a tool that is responsible for 35% of all the internet traffic in the world?
BitTorrent is a tool for file distribution which I wrote. It’s allows
people who are downloading something to automatically upload it at the
same time, so a very large piece of content can be delivered to a lot
of people without costing a huge amount of money to whoever made it
available initially. It works very well, so it’s wound up using a lot
How did you get people to use it in the beginning?
I made a useful tool and gave it away for free. It was handy enough to become popular.
Make any money off of it?
The company raised some investment money, and I’m currently employed working there.
How long have you been interested in puzzles?
I got obsessed with the Rubik’s Cube when I was in high school, and
figured out a solution to it on my own back then. It wasn’t a good
solution, and it took me several months of constant playing with it to
find, but for a pre-teen to find a solution without any help is pretty
For the Rubik’s cube aficionado’s out there how does your solution work?
I solve the bottom edges, then three of the bottom corners, then use
the missing bottom corner to quickly solve three of the middle layer
pieces, then do the last bottom corner, then use F-U2FU2RUR- to finish
the middle layer. Then I orient the top edges using FURU-R-F-, then
orient the top corners using RUR-URUUR-, then position the top edges
using R2B-RL-U2R-LF2BR2 and finally position the top corners by
keyholing using F-BF and its inverse. (When I teach people I show them
using FB2F, because there’s no inverse and alternating to learn with
How fast can you solve a Rubik’s cube?
A bit less than a minute and a half. I’m not particularly fast – I’m very bad at spotting the positions of pieces.
Was it your fascination with mathematics that brought you to puzzles?
Not really, I’d say the same thing which makes me interested in math
makes me interested in puzzles, although they aren’t directly related.
My familiarity with computational complexity gives some important
insights into the fundamental nature of certain kinds of puzzles
though, particularly burrs.
What kind of puzzles do you make?
My main interest started out being twisty puzzles, loosely defined as
‘puzzles like the Rubik’s Cube’. It generally includes puzzles in which
there are a bunch of identical pieces and several operations which
permute those pieces, and you scramble it and the puzzle is to
unscramble it. Puzzles of this form have a number of nice properties:
They’re always physically one piece so parts don’t get lost, they’re
much easier to unsolve than to solve, and it’s possible to invent one
without knowing how to solve it. (You can see videos of twisty puzzles here.)
I’ve more recently gotten interested in burrs, since I learned how to
use Burrtools. I’m too lazy to work on burrs unaided. I generally try
to make my burrs be all identical pieces, for aesthetic reasons, and
because it’s a criterion which forces me to actually do some work on
the puzzle myself instead of blindly giving Burrtools a set of
constraints and using whatever result it spits out. (This is an example of a burr puzzle.)
My other strange puzzle interest is puzzle rings. (Here is an example of a puzzle ring) Almost all puzzle
rings made for the last several hundred years have been the exact same
design, which is a very good design but there are lots of other
possibilities to explore. Puzzle rings turn out to be a very
distinctive genre of puzzle, which most closely resemble packing
puzzles in that they come apart easily and are hard to put back
together again, but they don’t come apart completely and have subtle
aspects of how the rings interlock which are most of the interest of
the puzzle and which there really isn’t an analogue to in any other
type of puzzle.
Occasionally I’ll do a puzzle which is in a genre other than one of
those, but my general approach is to take an interest in something and
go very deep in studying that one thing.
Most of my twisty puzzle inventions wind up being collaborations with
Oskar van Deventer, and he’s done a bunch of work on instantiating my
puzzle ring designs as well, which is why so many of my designs are on
Oskar’s shapeways shop. Burrtools is able to export directly to stl,
which is where all the files in my shop came from.
When designing a puzzle do you start at the end? or at the beginning? Do you start with the solution or the problem?
Your question presupposes a process where the puzzle inventor comes up
with a secret which is then hidden from the solver. Most of my designs
are so simple that it would be impossible to follow such a process -
the design constraints imposed by how it works make it challenging to
come up with a puzzle at all, much less one which follows very
preconceived notions of how it should work. It’s an extreme version of
form following function.
Explain the “form following function” issues behind puzzle design for us..
A lot of my interest in puzzles is in trying to explain to people that
even very simple motions in 3-space are much more mathematically
obscure than people think they are. Even Cartesian coordinates are a
hack which happen to provide a simple construction of 3-space, but wind
up making most of 3-space’s fundamental properties appear to be spooky
coincidences rather than being at the center of how it works. When I
say this usually people have no idea what I’m talking about, but when
people play with my puzzles the intuition gets across. Most of my
puzzle ideas are based on some notion of movement, and that function
almost entirely dictates the form of the puzzles, because 3d movement
has very strong mathematical underpinning which it’s impossible to make
small stylistic variants of.
What software do you use to make your puzzles?
Above is an image of the Ascii art, because our CMS is not equipped to handle it.
Could you recommend any of the tools you use?
Burrtools is an excellent tool for making burr puzzles, but it’s
extremely special purpose. Solidworks has a great history function,
which allows you to parametrize your model so that individual
dimensions can be altered and everything they interact with can be
automatically adjusted to go along with it, which can make some changes
be done in literally seconds which would otherwise require redoing the
whole thing from scratch. Rhino is good for some transforms done on
finished meshes which Solidworks is incapable of, like bending a helix.
Ascii art I can’t recommend in general
What brought you to Shapeways?
I think I first saw it when Oskar started posting things on it.
Why are you interested in 3D fabrication?
Many of my designs are so convoluted that it would be prohibitively
difficult to make them with any process other than 3d printing, and
even for ones which can be made in other ways, the amount of work
necessary to do multiple revisions or even sometimes to make a single
prototype is much less when done via 3d printing, and sometimes even
cost competitive if the pieces are made small.
Do you think that 3D printing will become mainstream or be for a niche?
3D printing as already fairly mainstream, in that it’s a common
practice to 3d print a master which is then cast afterwards for
commercial parts. If prices get down to somewhere around a fifth to a
tenth where they are now you’ll start seeing 3d printed parts as a
normal process for making consumer items, particularly if the Z axis
resolution is improved a lot.
The improvement of Z axis resolution(in layer resolution) is a good
point. Do you see 3D printing moving forward with everyone having their
own printer? Or would limitations such as this hold back such a
3d printers are expensive enough that it doesn’t make sense for people
to own them individually, since they mostly sit around unused and the amortized expenses are going up in smoke in that case (this is actually
true of 2d printers as well – those things are a such a ripoff). Moving
towards a person or company owning a printer and taking orders from
other people to do on it has been a huge improvement in how 3d printing
is done, so no, I don’t see anyone owning their own 3d printer any time
soon. If anything we’ll see an extension to the outsource model, where
companies own geographically distributed printers and are able to print
something physically close to you to have it ship faster.
If 3D printing does become a process to make consumer items which items
do you see it being used for? Do you see it changing society,
manufacturing etc. or will the impact be more limited?
I think 3d printing is likely to result in a wider variety of products,
made in shorter runs than they have been to date. There might also
start being customization of shapes as well as colors like we have
today, and possibly a lot of shapes in products which cause serious
difficulties for traditional manufacturing processes, although the
technical limitations of traditional methods aren’t all that onerous,
so I expect that to be a smaller phenomenon than one might hope.
Oskar informs me that you said the puzzle ring designs are all
printable in metal, and the price of the 6-banded design is about $20, which is
comparable to the cheapest puzzle rings bade with traditional
techniques and simpler designs, so it appears that there is now at
least one product for which 3d printing is just plain better and
cheaper. Puzzle rings are a bit exceptional though, as they involve
very little material, very precise workmanship, and have shapes which
are uniquely challenging. Any chance that white gold and platinum will
become available as printing materials?
Gold maybe, white gold less sure, platinum not in the near term.
What do you think of Shapeways what should we change, what should stay the same?
The process for people to order a piece if they aren’t uploading
anything needs to be streamlined a lot. Individual sellers need more
control over how their shops look and are organised, for example right
now if Oskar were to upload multiple sizes of one of the ring designs
they’d all be displayed on the top level and completely clutter up his
shop. There are also needs to be better SEO. For example, right now if
you make an improved version of a model you have to upload it
separately and take down the old version, completely breaking all
hyperlinks to the old one. Finally, Shapeways is currently very
unforgiving about requiring absolutely watertight models for printing,
and it would be nice if it did a reasonable job of accepting things
which were almost, but not quite, watertight.
We should be providing you guys with significant improvements to the organization of Shops and the multiple sizes in two weeks or so. The “versioning” issue is rather complex for us to solve but we are working on it. The watertight issue is also something we’re trying to improve.
You’ve said that “my general approach is to take an interest in
something and go very deep in studying that one thing.” Do you think
that this way of doing things is typical of an inventor? a coder? A
I think you have to do that to some extent to come up with an original
invention. It’s hard to come up with an original idea about something
if you just give it a cursory examination while other people have
thought about it deeply for a long time. The hardcore puzzle enthusiast
community tends to have broader interests, but I suppose that’s a bit
misleading since they’re vastly outnumbered by, for example, people who
are only interested in crosswords. Programmers used to all be widely
diversified, but that’s becoming less so since we now have clear
specialties in programming and languages which have meaningful
differences instead of just being worse versions of C.
Why is no one implementing IPv6?
There isn’t much clear benefit from it. What I’d really like to see is
for wireless router manufacturers to have UPnP port forwarding turned
on by default. Most of them have it built but turned off by default,
for no apparent reason.
With HTML5 will it be possible to use a code snippet that you can put
on any web page to turn that page into its own tracker, magnet URI, p2p
client, player the whole shebang?
Trackers need to have some persistent state, which I don’t think HTML5
provides. The embedded video functionality is interesting though, I
guess will soon find out if flash is more than just a video-playing gimmick.