[0] https://www.youtube.com/user/standupmaths [1] https://signalsandthreads.com/ [2] https://www.janestreet.com/puzzles/current-puzzle/
My impression with them in general was that they were willing to do lots of things that did not "conventionally" make sense at their size, and those things paid off. The internship program, for example, was relatively large in comparison to the size of the company at the time (≈50 interns?) and had lots of structure (events/talks/classes/group projects/etc) like you would expect from a big tech company, not from a 300-person firm.
They were also willing to build a lot of tools in-house like their own build system[1], their own code reviews system[2], etc. Most people would see this as wasteful NIH but I'm convinced it was a net benefit for them—they managed to be so productive in an absolute sense and especially on a per-engineer basis because they were willing to build so much themselves, not despite it. I'm sure the same thing applied to their recruiting efforts.
The biggest thing I took away from my internship was how much conventional wisdom in the software world was not necessary or true.
[1]: First Jenga, now Dune
[2]: Here's a neat talk on how they do code review at Jane Street: https://www.janestreet.com/tech-talks/janestreet-code-review...
It paid off for them, but I would say it's a risky move in general - it is very easy to get sidetracked with failed projects that have nothing to do with the business. There are many examples - Uber's internal chat system, etc.
Others tools were more a necessity, eg the OCaml build story was not great back then. Investing in OCaml itself was a calculated risk.
Also I wonder how things would be different if they were starting today, where there are more mature tools in the space.
Do you remember any specifics?
I’m also curious how their tooling made them so much more productive. Were the tools just really well designed, or did they integrate perfectly with each other?
And yeah, jane street is a pretty compelling demo that A) NIH syndrome is fine if you're good at writing software and B) it doesn't really matter that much if you use a mature language or some uncommon immature language
Because it is. What is the point of reinventing these wheels when gazillions of man hours have already been invested on open source tools that can do it better and cheaper?
2. Building something for yourself is qualitatively different than building something for somebody else. (I've heard this described as "situated software"[1].) Both the results and the process are different.
3. Building something yourself lets you become an expert in the domain and the tool you're building, often faster and deeper than using somebody else's system. It's a way to build up tacit knowledge and institutional capital as much as (or even more than) software.
4. More often than people realize, building something yourself ends up simply faster than first learning and then wrestling an existing tool into the exact shape you need. I've seen a lot of teams waste way more time trying to get some existing thing working than they would have spent building their own thing.
Obviously it isn't always true that building your own version of something makes sense, and nobody is going to be building everything from scratch... but it makes sense far more often than conventional wisdom dictates.
[1]: Term originally coined by Clay Shirky, but I can't find the original essay online, but looks like Gwern hosts a copy: https://gwern.net/doc/technology/2004-03-30-shirky-situateds...
A few factors not normally taken into consideration;
- Tooling is a great place to practice transferable skills on a practical problem. This always annoyed me with the XKCD cartoon 1205 which didn’t take into account improvement in optimization ability from practice in optimization.
- I’ve had a few external library dependency rug pulls, open source and proprietary. It never happens at a convenient time and is a total pain in the ass. Contractual agreements won’t save you unless you’re ready to sue for breach of contract and even that is no real solution. Perhaps getting code in escrow could be an alternative but I’ve never seen that work out either.
- External quality is mixed and getting worse. I’m careful with my own code so most of the bugs I hit is in other people’s code. If it’s a small library it can easily be less work for me to bring stuff in-house than to debug someone else's crap.
Honestly I wish I didn’t have to bring so much in-house, it would have saved me a huge amount of work. But we don’t have an efficient market with clearly defined standard of goods where we could treat software as a commodity and I’m not sure if we’ll ever get that.
Letting your development team control and build their own tools is a good idea, if they are able to. You probably don't want to do it in Java or C#, it'll take too much time, too many people, and be too unreliable.
Moral qualms aside, it must be a fascinating place to work.
> At the end of 2023. Jane Street employed 2,631 people, so that equates to almost $4mn of net revenue per head on average. In adjusted EBITDA, it comes to $2.83mn per employee (or nearly $22mn for each of the 482 traders actual traders at Jane Street.)
And regarding compensation:
> Given Jane Street’s disclosed compensation and benefits of $2.4bn last year, this works out to over $900k for each employee on average.
When I first heard of Jane Street, it sounded like Yaron Minsky was going around to MIT and such, high-touch, trying to hire just a few people. And later, things like this blog: https://blog.janestreet.com/author/yminsky/
The only negative thing I recall hearing is Jane Street alumni responsible for the infamous FTX and Alameda Research. I don't know whether the individuals were already fully into hopped-up sociopathic/narcissistic thinking in college, or whether their internship and employment experience contributed.
The things that made them great traders also made them much riskier choices to lead a larger company. To them, compliance is an obstacle in the way of profitable trades.
the bigger the vol, the more imma do
Is learning C++ a must?
I've been meaning to learn C++, but I always find it intimidating.
For quants (progression towards a trader / portfolio manager), Python / Matlab / R is enough.
> About 80 per cent of the company's capital comes from employee equity, which has swelled to $21.3bn at the end of 2023
o.O
I'd be interested to see if the Pareto distribution holds here as well, namely that 1% of employees (26) hold half the wealth ($10b).
20% of 20% of 20% (or 0.8%) of people will hold 80% of 80% of 80% (or 51.2%) of the wealth.
In binary, most significant(left most) bit is twice the magnitude of the previous bit. So that bit (or person) would contribute and get compensated accordingly.
Would also be similar to segment trees with each leaf node having a value of 1 with parents recursively summing the children?
> The real money is at the top. The bond prospectus reveals that Jane Street has 40 “equity unit holders on a full-time basis and in good standing”, with an average tenure of 16 years. Among those there will be at least a handful of billionaires, even if no Jane Streeter appears on any rich lists.
Sounds like any other partnership. A few people at the top are providing the equity and getting a profit share, and the thousands under them are getting salaries.
The difference is that these thousands also get to invest in Jane Street, which seems a pretty profitable investment (70% margins, etc).
Jane Street shares and profits are proportional to the capital you invest/accumulate.
How do you quantify, or even loosely, determine that?
What about ancillary workers who might not add any significant value to the (e.g. office janitor or HR) but supposing there’s a serious janitor or HR shortage then the org will still have to pay enough to attract someone, but what they pay is outside of their control and unrelated to the actual value to the company. And even in a worker-owned cooperative there’s still going to be in-groups and out-groups, and the in-group is incentivised to pay the out-group as little as possible as to maximise their own returns.
commodities trading houses tend to follow this model too though that is changing a bit.
i remember having this discussion with a friend after he sent me a richard wolff video. nothing about our system stops coops from flourishing. one of my favorite retailers, REI, is a member-owned co-op. publix, the beloved florida grocer, is employee-owned.
Finance tends to pay its workers better than shareholders—most banks’ trading and IB groups pay out more than 50% of profits to workers.
Where it goes wrong is when the regulator fails to stop foul play…
> A good mate who runs and wrecks expensive cars for hobby once reminded me that course plotting (i.e. research) and braking (i.e. risk management) are the two sine qua non contributors to a successful race.
Along with a reminder that disasters happen when businesses forget this (Boeing!)
Meanwhile I'd love to know what their edge is... It's probably more than OCaml, although... ;)
They are not making their money market making equity etfs afaik, mostly bonds. A slightly different game with a much higher barrier to entry etc
If you want to hire the best and willing to pay that is no longer a concern
They use OCaml because it is explainable. Which it is. They use it like a non-lazy version of Haskell -- side effects are used rarely if ever. So there's no nonlocal behavior in the code, which makes it easy to reason about. And that kinda matters when a lot of money is at stake.
Incidentally, Standard Chartered has their own compiler for Haskell, without the laziness. The group is led by the guy famous for Cayenne (the first dependently typed Haskell dialect).
Most companies don't even know that these are programming languages. Also don't assume people interested in these languages are "the best". Regarding JS, this is all hearsay, but I think nowadays they hire good profiles (competitive programmers / olympiad winner / top graduates from prestigious schools), ask them hard leetcode questions, and teach them OCaml (which anybody can learn, not particularly hard. Their talent pool isn't restricted to the OCaml community, as it used to be when they were less famous, except for niche use case (compiler work and so on...)
You got the direction the wrong way around.
Ask to introduce OCaml at any company. The first thing is they're worried about is not being able to find people. To which the answer is to train them. But then they're worried some people just won't be able to learn ocaml quickly enough. To which my answer is hire people who are good enough to learn it quickly. But then they get worried those people are expensive. Ultimately it comes down to not being willing to pay.
Yes, Jane Street uses OCaml, they have no reason to stop using OCaml, but may very well have been just as successful using another language. It's hard to tell, when we don't know the full circumstances of why they went with OCaml initially.
> but they could have been just as successful using another language.
His thesis is that finding one really good engineer in said language was 1 in 10 (10 interviews to find 1 really good engineer) whereas in more commonly used languages like Java, JavaScript, etc., it might be 1 in 100
[Edit] https://www.juxt.pro/blog/clojure-in-griffin/
If we had picked Python, it’s very boring and reliable, and the same could be said of Java. But you’re picking the lowest common denominator. I would say high performers, and the best programmers are often people that will only work in niche languages.
The problem is, there are good Java programmers, but there are also thousands of terrible Java programmers. If you pick the right niche, it’s easier to find the high-end talent. I think Paul Graham also made a very strong case that in a startup, you should be using the most powerful language you can, and that is Clojure.
I generally agree with the notion that talented folks are more likely to explore niche tech. Just be careful making the leap from "they prefer niche X" to "therefore they are talented".
Two anecdotes:
1. I'm an average programmer who likes niche tech.
2. My friend is 10x more talented than me, but he likes mainstream tech.
I don't claim to be a rockstar developer or anything close. But my capabilities and efficiency as a developer are tightly coupled to the tech stack I use (not just language).
I moved from a job where I chose my own tech stack that I iterated over several years to one where I'm forced to use (IMHO) tools that are poorly suited to my work, and I'd say the quality and volume of my work has dropped by at least 10x.
So I think it's both. You need smart people, but they also need to be using the right tools for the job.
However, I think if you are a company doing something boring and that can only pay average, then having an interesting tech stack (including a nice language), hiring globally and having good benefits might give you a venue to compete for talent. You'll need some kind of strategy.
The paradox being: developer familiar with programming languages of level of power N doesn't recognize that languages of level N+ are better (more powerful expressively), only that N- are lesser.
These days, starting or running a financial business with less popular languages is, well, less popular.
And one of the answers was Jane St. Apparently they produce great engineers.
Yes, I suppose this is all something to get all starry-eyed over, Jane Street encroaching on Citadel Securities, the two of whom control 30% of the US equity market volume.
I see it another way. I see people's hard earned money being siphoned by enormous financially-engineered vacuums, never to be seen again. And not just in the US, globally. This won't stop at 30% of the US equity market. It won't stop until the music stops and the last chair breaks. Which may or may not be soon. It will certainly be coming at some point.
Five times the London Stock Exchange’s entire trading volumes in 2023 in just your ETF arm? Sure ... this sounds like reasonable growth ...
> This is why some people argue that APs like Jane Street have become systemically important.
Oh, you don't say!
> About 80 per cent of the company’s capital comes from employee equity
That's adorable. They're like a little mom-and-pop shop ... except not anything like that.
can you expand on this? I have zero idea of what Jane street actually does and how they actually make money. (someone wrote that they have ~450 traders. trading what? equity? stocks? dark pools? PE? are they market makers? are they offering services to institution types?)
also what does "people's hard earned money" mean? you mean that Jane Street takes away their 401k or ... ?
We could essentially close them down if we moved all trading to say 1 hour a day.
speculators can then sell/buy to/from them and if they are doing it well, they make a profit. (and help narrow the spread.) sure, great, they even inject some liquidity. but why do we want a narrow spread? it only helps people who don't know which side of the trade they would rather be on, no?
We could essentially close them down if we moved all trading to say 1 hour a day.
Though most retail traders are upset they can only trade 7.5hrs a day, they want to trade 24/7 instead. They seem to WANT the liquidity and are apparently willing to pay for it.
Me, I'm fine with trading for only an hour a day, but I get not everyone is as lackadaisical as me.
Unless you have a better mouse trap to solve the liquidity problem, this is the best we have for now.
I actually know a little about this space. You know what the easiest response is that is always the answer?
"We provide liquidity".
Sounds important, most people don't get it, it works.
But it's not like the market is going to grind to a halt if Jane Street disappears overnight. Of course, people will say that. Not people telling you the truth.
I actually considered "but, but, THE LIQUIDITY!" in my answer but I felt there was sufficient snark.
That's funny, people say the same thing about ticket scalpers.
Artists set too low of prices because they don’t want to be thought of as greedy. Scalpers expose the true market rate of the goods.
No. Flow data is quite expensive, and most providers will only provide bucketed data unless you contribute with your own flows.
OP obviously has no idea what he's talking about but hey, what wouldnt you do for some internet points.
Envious bullshit! The reason they are so profitable is that believe it or not they are replacing earlier operators who were less efficient and taking more transactions costs out of the system before. To be anti-Jane Street is the same as being pro-Big Bank of the past, that was taking more money out of the economy doing a worse job!
Unless you know the history of global financial markets and lived it then you don’t understand. For example in the late 80s the CBOT treasury bond bit had over 1000 traders in that pit. They were all making money and quite a few a huge amount. And there was many more people supporting those traders of the floor. That was just a single futures bond contract! With Jane Street we have 2,631 employees doing the equivalent job globally and for less total cost of at least 80,000 employees in late 80s.
The profits per employee are higher obviously but that’s the effect of technology and productivity but the total price being charged to the economy as a whole is much lower. This trend will continue and I would not be surprised in 10 years that a company of 200 people will provide the entire function of those 2600 today, and probably the profits per employee will be $10M per person. But that’s what we want, is a good thing not bad, portraying otherwise is just Envy.
No, no ... it's not.
First of all I don't care one iota about Jane Street or how much money Jane Street employees make.
The problem is that this money has to come from somewhere. And while you say it's due to the entire system becoming more efficient, that's not the full story. That $1 billion Indian options play came at the expense of Indian retail investors.
So yes, there is a siphoning of money out of the hands of the retail investors and into the vast pool growing under Jane Street and its employees, much as it has done under others like Citadel.
Who is that good for, except Jane Street?
If you look the per capita income from where it's leaving to where it's going, that's where I see a growing problem.
There may be a sucker born every minute, but that doesn't mean that this is all just about making markets more efficient.
And, of course, none of this would have been possible without the additional "liquidity" provided ... the clearly essential contribution that makes it all justified. "We provide liquidity!". Ok ... ?
There's some efficiency improvements too, but ... eh.
Which means they're rapidly coming to a position which will easily allow them to game the system (what used to be known as "cornering the market").
I even wonder if their system has already learned cornering by itself via stochastic gradient descent.
These are flows, not ownership.
Also, taking Citadel as an example, a vast portion of these flows are not their own, but rather thirs parties offloading their flows to them, so they have the underlying best execution guarantee to provide.
An other way to look at it is that investors are moving away from traditional brokers to execute their flows, because these HFT firms have become so good. So instead, investors offload their flows to HFTs acting as DMM instead.
Removes any confusion for european and american collaboration.
Got old roots dating back to cockney trading slang: "The Old Lady just bought half a yard of cable"
Anyone aged around 60 or older would have been learning these numbers before 1974, and that's a significant overlap with FT's audience. You are dismissing this as if it's Middle English. It's an entirely reasonable explanation, at least one that shouldn't be dismissed.
The FT has a big international audience. As a German reader where a "Billion" is still 10^12 it does sometimes trip me up a little. So I at least find it useful.
So there was confusion, even in the English-speaking world. Was. If I understand correctly, England has adopted 10^9, and so now there is no ambiguity.
And instead of (metric) tons, I want megagrams. No kilotons; gigagrams.
I remember the first time I got those units out of gnu units on the command line scratching my head wondering why I'd never used them before and why it was using them at the same time.
The only 'commonly' used unit of measure larger than a kilometer is a light year, at least that I can think of. Maybe a Astronomical Unit, but not 'common' I suppose.
If the data consists of electrical potentials between 0.0000001 and 0.0002 V, it’s way easier to think about it as microvolts, especially if any other is in a complementary scale.
I've had people get angry when I tried to use Mm (mega metre) or ask why using some currencies symbols before the number when all symbols for all other units are used after.
Their effort definitely paid off though. Quants like JS, Citadel, or Jump hire some of the brightest students from Berkeley and other top CS schools.
B) if you have work experience, they filter on where you worked because the average google engineer / HFT engineer is probably better than your average engineer who works at missouri national bank.
C) if you've done something great that everyone knows about (like being the author of popular libraries, inventing tools that people use), then you can most likely bypass filters A and B if you can get in touch with a human recruiter (shouldn't be too hard).
For prestigious smaller companies which gets a lot of applicants, there is no easy way to reduce the pool without doing A and B. It is unfair and what college you go to might depend on circumstances beyond your control but if you have high ability regardless of what school you went to, you would eventually do B or C and get into the prestigious company.
If you don't have high ability and will never do either B or C, then they did the right thing filtering out your resume.
In the real world, it doesn't make sense for a company to discriminate against a candidate with higher ability just because they went to the wrong school (unless the company makes money from appearances like law firms, consulting firms or is corrupt and entrenched).
That being said, if you went to a community college but also had 10+ years experience at one of their competitors (eg: DE Shaw, 2Sig, Citadel, etc), you'd almost be guaranteed an interview. But once again, thats another very small set of applicants.
It wasn't a "logical way", it was an "easy way". That's classist at the very least.
I'm all ears if you have a better system that is more efficient. I'd also add that I went to a State college and my resume would have been tossed if I applied as a starting engineer too.
The question is what other attributes you "accidentally" filter by if you filter by school prestige. Even if you do bias your pool to individually better candidates on average, you may bias it in a lot of other ways.
If the purpose of the filtering is to reduce the pool to a size where humans with common sense can make sensible calls, then that CC graduate with 10+ years of competitor experience would likely be filtered out before common sense could suggest he might be good anyway.
HFT jobs are high value / high comp jobs, so Jane Street and their cohorts put a lot of effort into recruiting. This includes their resume intake system, internships with desirable colleges and professional outreach.
they get many applicants for a job. So you have to filter those applications. One way is to recruit from known places where they have already got people from as they know it worked. Typically those universities also filtered on how good you were and so the average quality is higher from those.
Obviously there will be exceptions but not work spending time to find them. Recruitment costs a lot in time and money.
I have applied about a dozen times since then after getting my degree from WGU and they haven’t gotten back to me; I almost wonder if not putting any college experience on the resume was a blind spot for their filter.
I learned math of finance from a professor I'm pretty sure was quanting on the Street at the time.
When I told him one of the constants in our little trading game was off by a factor of 10x, he said prove it. My final talk was: "How to make 2.3 million dollars in 30 days using options."
The real lesson wasn't about pricing options. It was about being ruthless when the time came.
You can't claim to have diversity if everyone did the same courses in the same schools.
That's a lot of money for someone who is essentially a middle-man. What a grift, about 13 dollars for each average American lost out to them a quarter.
Second, it's not at all obvious that market-making is a zero-sum game, where profits to the market maker are "lost out" to someone else. Market-making profits from the bid/ask spread, but a new market maker tends to reduce that spread. Ordinary people tend to invest as price-takers, and thus they benefit from a reduced bid/ask spread.
> We operate as a functionally-organized structure consisting of various management and risk committees. Each committee is responsible for directing the overall strategy of the firm and for emphasizing the importance of risk management to our operations. Each of our trading desks and business units is run by equity unit holders who take an active role in managing our day-to-day operations [...]. Our management structure allows for effective cross-departmental communication [...].
Yes, that org structure full of hierarchical business units with distinct responsibilities and committees and subcommittees and overseers and risk management totally screams "anarchy"...
> the organization of society on the basis of voluntary cooperation, without political institutions or hierarchical government; anarchism.
Maybe at the very top when you have more money than you’ll ever need, but for the vast majority of people jobs aren’t truly voluntary.
Because people seem to think it's "politics" and "government" that are the problem. It's not. It's people. It's always been people. It will always be people. People are the problem. Failure to acknowledge that people can and will be shitty is going to doom your system before it starts.
I'd say that Jane Street would seem to be less hierarchical than others and thus more closely aligned to an anarchist collaborative.
(It's all relative, I suppose. Bernie is a "leftist" in the US and would probably be a centrist in Scandinavia)
Was it an intentional rhetorical device to brush past the important distinction?
Yes, there will always be more powerful cliques. This is Jo Freeman's The Tyranny of Structurelessness. But the point is specifically whether those cliques and hierarchies are encoded into formal committees or whether they remain fluid based on the desires of the group at any given moment.
If we avoid codifying the hierarchies, we get better distribution of information and resources, people are held responsible to those who delegated tasks to them, and responsibility can be rotated more freely.
Sounds like a company.
Think when trying to translate a society to a corporation, then each of the terms must also be translated.
For a 'society' there is a government. And politics is dealing with the government policies, how we come up with rules.
But, if you want to translate to a corporation, you can't just say it doesn't have 'government and politics' because the corporation isn't dealing with the larger society. You need to also scale down the terms,
The CEO and the Hierarchy of bosses, are the 'Government' and 'political institutions' are the departments like HR, Marketing, and between them there are 'politics'.
There is no real 'anarchy' like individuals having freedom to 'just work on what they feel like'.
> Average TC of 900K
> Several unranked billionaires
it makes even OpenAI / Meta ML SWEs look underpaid
SWE who have deep subject matter experience are super valuable to these firms. Folks who understand how to write low latency code, FPGA work and other stuff like that. But the real money is in figuring how "how and what" to trade. Once that's done, the SWEs can bang out the code.
JS pays their new grads similarly, just a tiny bit lower.
New grad SWE is 400K
https://www.levels.fyi/companies/jane-street/salaries/softwa...
And levels.fyi not very accurate because 2nd year bonus is much larger than 1st year bonus in offer letter.
Quants/traders can hit $1M with 5 YOE not too difficult. Portfolio managers (similar to EM in tech) definitely $1M, sometimes $10M.
900k is average including janitors and HR.
Median for SWE most definitely over $1 million. Turnover of 6%? No way people stick around “only” making 500-600k new grad comp. Check levels.fyi
> Also, some high-level SWE do make 900K at Google and others
And some high-level SWEs at JS make 9 figures, not 900k.
Tether made 4.5bn net income in Q1 2024 w/ probably <50 employees.
Superfast arbitrage, which is Jane Street's main thing as far as I understand, doesn't really produce anything of value to humanity. Yeah, I guess the contributions to OCaml are worth something, but maybe not 10 billion?