Goldman Sachs v Russian Programmer

If you haven’t already read it, you should read Michael Lewis’ story on Vanity Fair about Goldman’s prosecution of Sergey Aleynikov in 2009. I recall the case when it first hit the wires, thinking that Goldman’s complaint that the code Sergey took when he left the firm could destabilize the financial system was incredibly silly. But like most people, I assumed that Sergey must have nicked Goldman’s HFT strategies and was guilty of theft.

That’s clearly what we were supposed to think. But in fact, Sergey wasn’t involved in the trading side of things at all. He was the brilliant back-end programmer who was brought in to speed up the uncompetitive latency of GS’s HFT systems.

There were many problems with Goldman’s system, in Serge’s view. It
wasn’t so much a system as an amalgamation. “The code-development practices in IDT were much more organized and up-to-date than at Goldman,” he says. Goldman had bought the core of its system nine years earlier in the acquisition of one of the early electronic-trading firms, called Hull Trading. The massive amounts of old software (Serge guessed that the entire platform had as many as 60 million lines of code in it) and nine years of fixes to it had created the computer equivalent of a giant rubber-band ball. When one of the rubber bands popped, Serge was expected to find it and fix it.

After studying its system, Sergey concluded it was such a mess that they should just scrap it and build something again from scratch, but his bosses vetoed that plan. So Sergey set out to fix the latency of GS’s systems by decentralizing it.

But most of his time was spent simply patching the old code. To do this he and the other Goldman programmers resorted, every day, to open-source software, available free to anyone for any purpose. The tools and components they used were not specifically designed for financial markets, but they could be adapted to repair Goldman’s plumbing.

So lots of little patches where put into GS’s code base, most of them hacked from code under an open source license, and GS’s managers would go in and strip OS licenses off and replace it with Goldman legalese.

Anyway, Sergey’s name starts to circulate around the street as one of the best HFT programmers in NYC, and he gets poached by a new hedge fund startup to build an HFT system from scratch. Double the sallary and the opportunity to do something cooler than constantly fix Goldman’s dogfood system.

Before leaving, he uploads a bunch of code to a subversion repository. Not the strats–he didn’t work with those–just a collection of patches and fixes that he used, most of it OS code taken from the internet. It was like taking a notebook, a collection of reminders for solving IT plumbing problems. It was of no use to anybody else and certainly wasn’t Goldman’s secret sauce. Arguably, it wasn’t even Goldman’s property.

What transpires is a total miscarriage of justice. One can’t help but speculate that GS’s motives were to use Sergey’s prosecution to create the impression that its systems were better than they actually were. Who knows. Read the whole story.

Healthcare Costs and Technology

There is an excellent article in the MIT Technology Review this month (The Costly Paradox of Health-Care Technology) pointing out how healthcare is the only industry where technological progress appears to raise rather than lower costs.

The reasons are not a mystery:

Unlike many countries, the U.S. pays for nearly any technology (and
at nearly any price) without regard to economic value. This is why, since 1980, health-care spending as a percentage of gross domestic product has grown nearly three times as rapidly in the United States as it has in other developed countries, while the nation has lagged behind in life-expectancy gains.

Other researchers have found that just 0.5 percent of studies on new medical technologies evaluated those that work just as well as existing ones but cost less. The nearly complete isolation of both physicians and patients from the actual prices paid for treatments ensures a barren ground for these types of ideas. Why should a patient, fully covered by health insurance, worry about whether that expensive hip implant is really any better than the alternative
costing half as much? And for that matter, physicians rarely if ever know the cost of what they prescribe—and are often shocked when they do find out.

Yet the article concludes when some policy recommendations that range from vague (organisational change, innovations in health care delivery) to downright dumb (“drug container caps with motion detectors that let a nurse know when the patient hasn’t taken the daily dose.”).

The solution, as I seed it, is straightforward: health insurance that pays out a lump sum of cash per diagnosis, to be spent however the patient sees fit (some sort of trust/trustee mechanism needs to exist for those too ill to make the decision themselves). The current framework, whereby insurance pays for whatever treatment doctor thinks best, provides absolutely no incentive to make the inevitable tradeoffs between cost and expected benefit.

Perhaps when healthcare inflation eventually leads to rationing, patients in America will reconsider the wisdom of this paternalistic model and demand the right to make those decisions themselves.

Big Data and Price Discrimination

There is an article in Forbes on how big data is brining about more first degree price discrimination. It summerises a recent paper by Reed Shiller at Brandeis on the subject, who studied Netflix’s pricing.

Simulations show using demographics alone to tailor prices raises
profits by 0.14%. Including web browsing data increases profits by much more, 1.4%, increasingly the appeal of tailored pricing, and resulting in some consumers paying twice as much as others do for the exact same product.

Even the price is tailor-made for you, Sir.


I’m a big fan of this concept, so this bit of anecdotal evidence is discouraging. A BBC reporter spent a week trying to earn a living from crowdwork. In total: 37 hours worked, 19.16 GBP earned. I’m sure others can do better (and this was only an experiment), but still.. does anyone know of some proper stats on the hourly pay distribution of crowdsourced work?

I can’t help but think that these skills would be renumerated better if they were hired as employees or contractors. If that’s correct, I have some guesses as to why:

  • Buyers of the skills get less value from the workers because of the quasi-anonymity and one-off nature of the relationship.
  • The middle-man doesn’t add much value.

In light of the news of Ronald Coase’s recent passing away, we might want to consider whether this crowdworking phenomena can be understood in terms of his transaction cost analysis of firms.

I’ve had some limited experience as a consumer of crowdwork. My impression so far is that you have to wade through a sea of rubbish before you find someone worth paying. And when you’re there, you’d really rather deal with the person one-on-one in an on-going, no-commitments basis. What makes all of this work is search and reputation, and a fragmented hodge podge of different crowdwork platforms doesn’t really perform either of those functions very well.

Indeed, the very term “crowdwork” suggests the wrong framework, in my opinion. Most work takes place over time and needs to be integrated by the entrepreneur or manager with other work to make the whole. Both of those factors require lots of tacit knowledge on the part of both worker and employer. Tacit knowledge is the sort of knowledge that only really exists in the minds of individuals or small groups. You can’t share it with a paid “crowd”.

To make this empowering vision work, I think we need a protocol rather than crowdworking platforms.

Oh, and the highest paying gig the journo did was.. getting paid to click “likes” on websites, something that requires no skill at all!

What’s News in The Classics

Ancient Egypt

The BBC has this story covering a Royal Society paper: New timeline for origin of ancient Egypt.

Radiocarbon dating suggests that Egyption civilisation came into
being much later than previously thought.

Previous records suggested the pre-Dynastic period, a time when early
groups began to settle along the Nile and farm the land, began in
4000BC. But the new analysis revealed this process started later,
between 3700 or 3600BC.

The Palermo Stone is inscribed with the names of early Egyptian kings
The team found that just a few hundred years later, by about 3100BC,
society had transformed to one ruled by a king.

So the pre-Dynastic period where people began to settle along the Nile and become agricultural transformed into a society with a state and single king in the short space of a few hundred years, much more rapid than historians thought. Very interesting stuff.


Mary Beard reminds us that smear campaigns can lest well into posterity. On Caligula:

But even the more extravagant later accounts – for example the
gossipy biography of Caligula by Suetonius, written about 80 years
after his death – are not quite as extravagant as they seem.

If you read them carefully, time and again, you discover that they
aren’t reporting what Caligula actually did, but what people said he
did, or said he planned to do.

It was only hearsay that the emperor’s granny had once found him in
bed with his favourite sister. And no Roman writer, so far as we
know, ever said that he made his horse a consul. All they said was
that people said that he planned to make his horse a consul.

The most likely explanation is that the whole horse/consul story goes
back to one of those bantering jokes. My own best guess would be that
the exasperated emperor one day taunted the aristocracy by saying
something along the lines of: “You guys are all so hopeless that I
might as well make my horse a consul!”

And from some such quip, that particular story of the emperor’s
madness was born.

It’s a short piece and worth reading (sorry, it’s a month old.. but hey, this is ancient history).

Classics in East London

This week’s Economist has a piece “Latin, innit” (it’s behind a paywall) about BSIX, a sixth form college in Hackney, a poor part of East London. Unusually, they have a classics programme with 17 enthusiastic students.

Several students say they plan to apply to Oxford. And on August
23rd, the East End Classics Centre ws given some money from London’s
Schools Excellence Fund to expand and link with other similiar
projects. In time, it may seem odd that the sex and violence of the
ancient world were ever absent from the classrooms of London’s East

I really hope there is more of this. It is a scandel that the subject is viewed as an irrelevant past time of the upper classes, a stupid misconception popularised by Labour politicians that only holds back the very people they claim to represent.

Outside Perspectives

It’s always enlightening to get an outside perspective on one’s country and times.

A Pole wakes from a 19-year coma to find the Communists ousted from power and no more petrol queues:

An Indian international student’s observations on America:

After reading that, you will enjoy watching this:

The Economics of On-line Learning

Virtual Growth Projects has an interesting essay on the economics of on-line learning. The essence of the idea is that the teacher/student relationship in the on-line world has allot in common with the economics of file sharing:

The seed/leech terminology is borrowed from file-sharing where a
‘seeder’ is someone who possesses 100% of a file and is in a position
to share it with others. A ‘leech’ is someone who does not yet have
the file, and hence cannot share it.

  • a knowledge-leech has not yet achieved mastery of the subject
  • a knowledge-seeder has achieved mastery (and hence is in a position to share it with others)
  • knowledge transfer is converting leeches to seeders

The knowledge-transfer process creates ‘seeders’ who are capable of
productively contributing to it.

Given that the best way to demonstrate understanding of a subject is the ability to teach it to someone else, it seems plausible that the business model that on-line education is likely to gravitate towards is one where students evolve into roles involving teaching, essay grading, exam question writing, etc as they gain mastery of the subject that they are learning. Success in your studies would, in this model, act as a sort of currency that would allow one to recoup some or all of the up-front costs involved in taking a course.

It might also take us away from the traditional production model of pedagogical contents–expert writes a text book and course materials–to one of edited but crowd-sourced content, much like Wikipedia.