Start Ups, Crypto Companies, and Commercialization

[Tim May]

On Sunday, December 23, 2001, at 01:29 AM, Bill Stewart wrote:

> At 02:19 PM 12/22/2001 -0800, Tim May wrote:
>> Yes. The best work has always been done by one or two people at a time.
>> This applies to software as well.  (Not so much to chips anymore,
>> at least not for the past 20 years. Another topic.)
>
>> A person with the dedication and skill of a Stallman could
>> probably implement digital cash without having the Herman Miller 
>> chairs,
>> the hot tub up on the roof of the office building, the staff of 
>> marketdroids, and the espresso machines.
>>
>> There's some hope.
>
> Espresso is important, basic machines are cheap, and the real decision
> is whether walking down to Starbucks to avoid making it yourself
> is more an interruption to your concentration or an opportunity
> to spend time in the real world and to check out the Mondex smartcard 
> machine
> that Starbucks gave up on using :-)

I was using that as a stand-in, a place holder for the recent trend for 
start-up companies to spend lavishly on corporate digs.

Being an old-timer, I like to say: In my day, we had to walk five miles 
through the snow to get a cup of mud from the vending machine. Actually, 
in my day at Intel we were lucky to have patty melts for lunch, as most 
of us ate out of vending machines (burritos, stale sandwiches, cottage 
cheese) or out of the vending vans ("roach coaches") which pulled up 
outside to feed the engineers and operators.

Things have changed, I gather.

I've got nothing against companies having nice facilitites, but perhaps 
the lavish cafeterias, rooftop jacuzzis, and other frivolities give the 
workers the impression that money is flowing freely (as it is...for a 
while). I'd rather see a company starting on a shoestring, with surplus 
desks from Repo Depot, with crowded offices, and with no money wasted on 
frivolities. Especially if I'm an investor!

Sure, it'll cause some of the folks to go instead to Oracle, where the 
cafeterias are opulent. A good thing that those who want opulence go to 
such places!

(I faced a similar choice in 1974: some of the places I was considering 
were known for their lavish expense accounts, their nice offices. Other 
places were more spartan (that's a good word to think about). I chose 
one of the spartan places, Intel. A lot of those other companies I could 
have gone to went nowhere.)

This applies to marketdroids as well. Startup companies get larded-up 
with marketing departments when there is little to market. The huges 
staffs of some of the companies we all know about is an example. (Just 
as bad: the hiring of ex-government officials, regulators, etc. But I 
digress.)

>
> But the easy part of doing digital cash is the software,
> and it doesn't take years of Stallman-level or Chaum-level or
> Ian-or-Ben-or-Lucky-level wizardry to produce it, though it's
> really helpful to have their insights into what didn't work
> and what pieces were useful for consumer-quality realizability.
>
> Lots of people can turn the algorithms into reliable code;
> lots of people can build user interfaces, though you if you want
> it to run on the latest Microsoft GUI API environments and
> all Mac environments from 6.5 through 10.1.2Coca you'll need a few
> extra helpers to add the ugly details.  (*I* could even do the 
> programming,
> though you'd get a basic web forms interface and a text interface
> that looks suspiciously like "throw stone knife at dwarf", "404 Knife 
> Not Found",
> with none of that modern Javascruft or "ncurses" aesthetics in it :-)

I agree and disagree. I agree that the core crypto is sort of 
established. However, how many examples of it do we see? Few. How many 
of the programmers here on this list (there must be a few dozen who call 
themselves programmers, professionally) have ever implemented anything 
remotely similar to digital cash?

(I don't call myself a programmer, but I fool around. The closest I have 
come to the above is when about 10 years ago I wrote a simple RSA 
implementation in Mathematica, just to make sure I had all the "Euler 
totient function" kind of crapola down straight.)
>
> The hard part is getting people to take the stuff.
> 30 years of Kernighan, Ritchie, and Thompson, 20 years of Stallman and 
> Gilmore,
> 15 years of X and Gosling, and 10 years of PGP & Linux has gotten us 
> partway to
> World Domination in technical areas, often by getting the good parts 
> stolen badly
> by the bigger commercial interests, but money's harder to change;

PayPal took off pretty fast. So did VISA and MasterCard in their day. I 
think and hope someone will cut through all of the b.s. and do something 
that takes off.

As I mentioned a few days ago, one of the reasons we wanted to have a 
session on "implementing digital money"  was to brainstorm these issues. 
Maybe we should try again.

By the way, if "lots of people can turn the algorithm into reliable 
code," where _are_ the implementations? I see bits and pieces.



> Black & Scholes and Fair & Isaac and Milken and Visa/MC/AX and Schwab 
> and
> later E-Trade and just possibly PayPal have changed things,
> but Mondex didn't happen, and Micropayments didn't happen,
> and in spite of all of Hettinga's enthusiam and Chaum's business acumen,
> anonymous digital bearer cash hasn't successfully rocked the world.
> It not only takes technical skills to ship working stuff,
> it takes business skills to find a market where it works and
> promote it enough that enough people are using it that
> some level of anonymity can actually happen.

I don't think things happen because of "evangelists." Evangelists didn't 
give us the transistor, the IC, the microprocessor, the early personal 
computers, or even the Mac (where Apple was famous for hiring 
"evangelists," e.g., Guy Kawasaki).

Discussing this would require a longer article here than I'm interested 
in writing. To be sure, advertising BankAmericard (VISA) and other 
credit cards was and is a big business. People have to know a technology 
exists and then want to use it.

Evangelizing digital cash, when no real digital cash implementation 
exists, is getting things backwards.

> Lucky and the Mark Twain Bank had the technology, and had the
> service working in the abstract, aside from the minor problem that
> there was nothing to buy except pictures of Cypherella
> before they stopped allowing that, though perhaps if they'd
> been a bit later to market and jumped into E-Bay when PayPal did
> they could have pulled it off (or perhaps not - that's a market
> where reputations have a really high value, and you'd have to
> structure an escrow market around your digicash that would
> undo most of the anonymity even if the digicash provided it.)

Lucky can tell us what the real level of technology was. My impression 
is that it was a cleaned-up version of Chaum's earlier code (or the code 
of his early 90s programmers, that is) and did "uninteresting" things.

And since it wasn't payee-untraceable, interesting uses for trading 
banned materials were not possible. (My recollection, though I could be 
wrong, was that Mark Twain Bank also had the usual ISP-like junk about 
acceptable uses and how accounts could be cancelled for "inappropriate 
uses.")

A lot of these applications are just "toy uses." Not even the True 
Believers, most of us, would waste our time and money opening a Mark 
Twain Bank account so we could flash our account cards, or whatever, at 
local parties and meetings. I don't know of anyone, besides Lucky, who 
ever used the system.

It is true, and we've talked about it many times, that most people don't 
care about anonymity and untraceability. They don't care that their 
"Fast Pass" turnpike passes can (and sometimes are) be used to track 
their movements, to find out when they were on the New Jersey Turnpike 
and which exit they got off at. 99.9984% of them figure no one will 
bother to check. So the market for Chaumian ecash for car passes never 
materialized.

A big part of the problem is the lack of evolutionary learning. This is 
a problem with our patent system, especially for software. A number of 
years ago I wrote an article about how patents for hardware work because 
the produced good "meters" the patent: no one cared what uses were made 
of the microprocessor because every sale was a sale, thus paying for all 
of the various R&D and patents and so for the chip companies. With 
software (*), there is much concern about what uses are made. Because of 
replicability of the product.

(* I don't mean a software product like Microsoft Office. It is true 
that no one cares what use is made of a copy of Office, provided it was 
legitimately bought. I mean software like "RSA," where sale of a general 
license has to be very carefully planned.)

Because Chaum wished to make money (not an ignoble goal), he limited 
access to the core of his system. (Whatever you want to call it, the 
algorithms, the implementations, the ability of others to build 
products, etc.) There were a series of "future by design" projects, but 
little evolutionary learning. (Contrast to the aircraft or chip 
businesses, where hundreds of companies failed, planes crashed, new 
designs were tried, patents were cross-licensed or bought, companies 
rose and fell, products proliferated...)

It's not easy to build a company around an algorithm. RSA succeeded, but 
it faced years of shoestring operation troubles (I visited their crowded 
offices in Redwood Shores, circa 1990-1). And it had arguably the most 
important patent portfolio of all. (Levy's "Crypto" details the history, 
and the almost out of business experiences.)

I see way too many Cypherpunks jabbering about "raising money." Most 
don't.

It's time for a return to the older models.
>

> Doug and the Austin Cypherpunks Credit Union folks had the technical 
> skills,
> and the interesting hook that in the US, Credit Unions have
> much less regulation than Real Banks, but figuring out how to make money
> from such an activity was tough, and unlike MTB, they decided not
> to launch a business they didn't know how to make money with :-)

Good for them. Both later made some money from other projects.
>

> Getting the real thing working requires real marketing skills

See above. I doubt a marketing group makes a hill of beans' worth of 
difference.


> and being in the right place at the right time; occasionally

True enough.

> you can hit it off, like the kid who wrote WinAmp and was pressured
> by his parents into making it Shareware and not just freeware,
> or the Hotmail folks causing the free-web-based-email wave
> (and catalyzing many of the appallingly stupid Dot-Com Business Plans.)

We saw a lot of appallingly-stupid Cypherpunks Business Plans, too.

"And then at this point the world adopts the use of Bearer Bucks (TM) 
and we get 2% of every transaction!"
>

> Perhaps one advantage of the dot-com crash is that people starting
> businesses today are much more likely to do the solid business planning
> and the initial technical decisions before they get enough
> funding to leave the garage and hire the 200 programmers that
> it takes to prevent any real work from being done while you're
> having meetings to coordinate development of the hot-tub-scheduling 
> website.

Yep. Premature commercialization, mythical man-month, burn rates, and 
all that.

>
> But if you're not going to use the marketdroids, you have to find
> some really solid alternative to get the stuff widely used.

I'll say it again: the best commercialization is done for stolen 
products.

(Or, to head off frivolous charges that I am libeling Sun or Cisco, 
below, products which were basically already developed and faced a ready 
market.)

The Stanford University Network (SUN) machines were largely ready to go 
when the founders of Sun (gee, where'd they get that name?) sought 
permission (and investment) from Stanford to commercialize them. Ditto 
for Len and Sandy's work at Stanford on routers. This is how Sun and 
Cisco were able to "hit the ground running." They didn't squander 
precious startup dollars on roof-top jacuzzis and meeting rooms for 
researchers to sit around trying to develop a product. Investment money 
went into production facilities, wire-wrap guns, etc.

Even RSA was basically just commercializing an already-extant thing.

Intel was making and selling actual products within 6 months of its 
formation. (It did _not_ get funding and then sit around in opulent 
surroundings thinking about future directions. In a way, it also "stole" 
its technology, in that its founders had worked on silicon-gate MOS at 
Fairchild and knew how to get rolling quickly. I'm not accusing Intel of 
stealing in any prosecutable sense, but in the sense that this is the 
way evolutionary learning happens: the children of successes create more 
successes.)

By contrast, I've watched dozens of companies (some of them started and 
staffed by my friends) raise some seed capital and _then_ begin their 
research and development! Bad move. For lots of reasons I could write 
about for hours. (Xanadu and AMIX were examples, several Cypherpunks 
startups are other examples.)

(Nutshell: research proceeds unevenly. Breakthroughs. Evolutionary 
learning. Dead-ends. Redirections. "Exploitation of rich veins of ore," 
punctuated equilibrium, time value of money. When a company gets 
investment money, it _must_ begin to use that money _immediately_. 
Ideally, for production facilities, for actual product advertising, for 
_immediate_ uses. The time value of money dictates this. It _cannot_ use 
its money to hire people to sit around and think about future research 
directions. If there is not a _real_ business plan (I don't mean a 
"spreadsheet business plan") with a real revenue stream beginning 
_soon_, why form a company? A lot of naive people seem to think the 
capital markets exist to help them fund their dreams, their vague ideas 
about becoming the next Bill Gates. Maybe the recent dot com collapse 
has changed things.)

There _is_ a place for "service businesses." ISPs, for example. (The 
founders of ZKS made their initial money by setting up a successful ISP 
in Canada and then selling it when the great wave of ISP consolidations 
was cresting. Interestingly, some of the early remailer operators in 
Holland did something very similar, selling their ISP business. Looks 
like this was the real way to make some money. No exotic cutting-edge 
technology, just paying customers.)

There's a saying that "the best is the enemy of the good."  This is part 
of the long debate in computer science between "doing the right thing" 
and "doing what works." The "Right Thing" is the elegant, the 
crystalline, the pure. "What Works" is the crufty, the cobbled-together. 
LISP versus Perl, perhaps. Richard Gabriel has some fun essays along the 
lines of "Worst is Best" (use Google to find them), discussing why the 
workstation and language market developed in the way it did.

In our community (and related orbiting communities), "the cool is the 
enemy of the good." Hence we see multi-year efforts to the really cool 
implementation of hypertext developed, thus running out of money and 
missing the boat on the Web. (Xanadu.) Ditto for digital cash, as 
comparatively low-tech and uncool products like PayPal move in to take 
the low-hanging fruit.

Paying people to sit around and dream about rilly, rilly cool products 
is the kiss of death.

Lastly, the focus on commercialization has been very weird these past 
several years. Nearly everyone jabbers about how to make money off of 
remailers, or data havens, or digital money. But little in the way of 
new ideas are being discussed here.

Our last Cypherpunks meeting was refreshing in that we had a couple of 
good talks on real things and very little of the "lawyer" and "startup 
evangelizing" junk. And we ended with a heated discussion of remailers 
and such, which Len and Steve and others have been writing about.

Even more lastly, you folks out there thinking about how to do a startup 
or make money should be thinking about what I said, and which I am 
certain is basically correct: start up companies are not the place to do 
basic R&D. That's the place for universities and for established 
companies (with the companies either spinning-off divisions or selling 
the products themselves or losing their staff who "steal" the work).

The cost of money is just too high for anyone to fund blue sky dreamers. 
Except as charity.

(I can't resist: Look at Interval Research. Paul Allen, who has almost 
as much money as Bill Gates, decided to try to replicate the success of 
Xerox PARC and funded Interval Research some years ago (around '92-'93, 
IIRC). He gave it a wad of money, they set up shop, they hired Brenda 
Laurel and Lee Felsenstein and all kinds of other bright thinkers and 
dreamers. What came out of it? Nothing. Nada. Nil.)

--Tim May
"As my father told me long ago, the objective is not to convince someone
  with your arguments but to provide the arguments with which he later
  convinces himself." -- David Friedman