bfrink

You hit HELP HELP (a dedicated button on the keyboard) and a real, knowledgeable person in your time zone will _fix_ any problem with the data you see on your screen. If it's not fixed by the time their day ends, they will hand your ticket to yet another knowledgable person to the west who will continue working on it to fix it. That is, the data quality, and the data quality processes, are a large part of what you pay for. People get hung up about real time, but what if a number on a financial statement from 10 years ago is screwing up your model training?

The US publishes financial statements for free on EDGAR. But then you have to parse the unstructured data, normalize it so you can compare a bank to an airline to a manufacturer and then you have to accommodate mergers and acquisitions throughout history and every revision a company makes for prior periods, and then you're starting to talk about real money and the price of Cap IQ or Compustat or Factset starts to look pretty attractive, especially if your alpha doesn't come from your proprietary methods of databasing financial statements.

Y latency doesn't matter if a human is looking at something. By the time your caveman-speed brain reacts, an FPGA sitting in an Equinix data center has made a hundred trades.

Seems that something's been broken. The normal way to look up ticket images manually at https://wmq.etimspayments.com/ no longer works, returning "Unauthorized" even after properly getting past the CAPTCHA.

Indirect rates are negotiated. What are the incentives for the government negotiators to get the lowest possible rate? I honestly don't know; I'd like to understand more about the underlying drivers here.

Programs like Stars College Network (https://starscollegenetwork.org/) and Questbridge (https://www.questbridge.org/) help to bridge this gap in knowledge. They are really good programs, based on my limited to exposure to them as a Caltech alumnus. It was an incredible stroke of luck that I knew Caltech even existed growing up in a very small town pre-Internet, and these programs take some of that luck out of the equation.

Interesting - I have a contact in the CTLO (the real teaching-oriented group) and they have said they have no contact with the CTME and are discouraged from interacting with them.

While the provost may have approved it, this whole thing is another example of administrators running roughshod over faculty (who are, admittedly, disinterested in this sort of thing). I do (hazily) recall that back when I was a frosh, there was some sort of professional program that focused on engineering management and was a joint effort with places like Hughes and The Aerospace Corporation. Given Caltech's long history with the Southern California aerospace industry, this is at least plausible. This was done through the Industrial Relations Center (page 21 here: https://campuspubs.library.caltech.edu/121/1/1992-1993.pdf)

I agree wrt outreach. It should be Watson lectures and setting up telescopes, etc. This isn't outreach, is an obvious money grab.

Agree on the sensitivity wrt traditional (East Coast) media. Imagine what it takes for the president (and provost, and general counsel, and chief communications officer, and ...) of Caltech to send out an email at 8:30pm on a Sunday night. This article (finally) rattled something at the Institute. However, they are still defending this program, in court, and in the press, as a wonderful thing Caltech is proud to put its name on. 4.6 out of 5 stars! (How many NSF and NIH grants get funded if they are rated 4.6/5 by reviewers? Hint: zero)

You've got faculty calling it a major embarrassment[0] and alumni aligned 100% against this grift. One has to ask, if faculty and alumni (and I presume current students) don't want this program, then who is advocating for it? One can only assume the worst.

[0] https://x.com/CFCamerer/status/1840545142006288667

32% of 2023 graduates had any debt at all; the average total indebtedness (over all four (or more) years) of that 32% was $15,896[0]. Not bad for a school that costs $86,886 _per year_[1].

One should also note that "[s]tudents from families making $90K or below, will receive a no loan financial aid package (package will consist of grants and work-study)" and "[a]t least 20% of first-year students in the past three incoming class were Pell eligible students." [2]

[0] https://www.finaid.caltech.edu/typesofaid/loans/disclosure

[1] https://www.finaid.caltech.edu/documents/25301/2324_tuitionf...

[2] https://www.admissions.caltech.edu/afford

Caltech accepts a tiny, tiny amount of transfers in total; less than ten per year. The imbalance in transfer admissions is not significant.

This (seemingly elegant) Virginia-class implementation feels like an inheritance of the Rickover and SUBSAFE philosophy. Not sure how the surface Navy would accept anything less.

At Caltech, students from families making $90K or below, will receive a no loan financial aid package (package will consist of grants and work-study).[0]

[0] https://www.admissions.caltech.edu/afford

A large Great Lakes cargo vessel was recently completed by Fincantieri Bay Shipbuilding (https://en.wikipedia.org/wiki/MV_Mark_W._Barker). Great Lakes vessels need to be built in the US due to the Jones Act, but they also tend to last a very long time in the fresh water in which they operate, so this is the first lake freighter built in many decades.

Ships for the US Navy (including the Military Sealift Command) are also built in the US. So, shipbuilding hasn't stopped, but the US's comparative advantage is not really in shipbuilding.

You mean like this billionaire philanthropist? https://chanzuckerberg.com/science/programs-resources/open-s...

Value investing like Fama French HML? Value like the factor that's gotten trounced over the last decade?

Agree - real world discrete things should be modeled as such. If MPV was $0.23, then model $0.23 increments - whether you use fixed point, or the cardinality of increments, who cares. But all the other math leading up to a discrete decision on the increment is almost certain to be best described with, and faster to implement in, floats.

In your example, you're only a penny off if you truncate 0.199999999999999996, rather than rounding (which is described in IEEE 754!). Here's a real simple example. Let's say your model depends on the average of the last three ticks. The last three ticks are $1.00, $1.00, and $2.00. Ok, what's the (exact!) average without being off by a fraction of a penny? This is the point - as soon as you start manipulating numbers in anything other than the most trivial way, you run into the dreaded floating point error, because that's how the real numbers work.

I am unaware of fixed precision types that have hardware optimized (other than FPGAs which are used for feed handling in HFT anyway). If you are modeling discrete things like Minimum Price Variations, then yes, use fixed precision, or even encode it in a way that saves space. But if you're numerically solving a partial differential equation, e.g., Black Scholes, it's difficult to see how fixed precision numbers are going to have an advantage.

This is a common misapprehension. Yes, fixed precision is great for accounting (like your bank statement) but when you're making a predictive model of asset prices (or higher moments of price distributions) being off by 1 in 10^14 is not important (because your model isn't that precise anyway!) and the performance you get from dedicated floating part hardware is well worth the "loss" of precision.

This wholesale dismissal of floating point for "financial" systems ignores the real business needs which might point you toward fixed or floating point numbers. Always ask yourself - do you need to know this number to better than 1 in 10^14? Are you going to find its square root at some point? Also remember that storing fixed point numbers usually takes more bytes than double-precision floats.

Calcbench (https://www.calcbench.com/) does this, but as others point out, XBRL is a bit of a mess

There are about 30,000 institutionally-investable stocks in the world. They have ~252 closing prices per year. This is the thing you're trying to predict, so it forms the outputs of your training set (be sure to hold some back for out-of-sample testing and there are probably distinct regimes you need to switch between). Yes, there may be lots of features (weather, geo populations, etc., as you point out), but you're often hard up against the curse of dimensionality here. Sure, intraday data is more voluminous, but there are fewer economically reasonable influences on that.

I think there are other wealth management-related startups that are actually changing the paradigm, e.g., Wealthfront, whereas (to the extent of my knowledge) Addepar is helping the incumbents in the space continue to capture fees in excess of their value add (your original point). Relative to other local firms, Addepar operates at an information asymmetry advantage - the finance and especially wealth management industry is less well understood by the average CMU SCS graduate than, say, the social media industry. I think Addepar exploits this.

Of course everyone starts smaller than their end state, but as far as I can tell, Addepar still focuses on client reporting. There's a lot of data aggregation, etc., that goes into that, but from their website: "Addepar gives you a competitive advantage. By eliminating the manual burden of aggregating your financial information, and making it easy to generate customized reports in just seconds, we free you up to spend more time designing and advising on client investment strategies." If there's more there, like portfolio construction/rebalancing, order generation and execution management, clearing and settlement, etc., it's not easy to apprehend from publicly available information.

Look, I think Addepar is amazingly beautiful software, exceedingly well executed (though it may be pearls before swine.) It also has tremendous hype (which is often seen as an unalloyed good in the Valley). Is it a viable business? Is it revolutionary, or is it a much-better executed Advent with the advantages of no legacy code- and userbase? I think those are interesting questions.

I think perhaps the worst thing about Addepar is the way it sells itself to prospective (inevitably young) employees: that's it's on a mission to "fix finance." While there are some operational inefficiencies to be alleviated in the wealth management performance reporting space, the savings from which might at some point be passed along to asset owners, the actual result of Addepar's work, at least in the short term, is much less grandiose. I would characterize Addepar's effects as enabling wealth managers to continue to capture more of this value as you point out, while also assuring tax-efficient inter-generational wealth transfer.

This is perhaps a cynical and short-sighted view of Addepar (and I've been told as much by Addepar's management), but based on my experience in the investment management industry, I feel it's more true than false.

Kaiser Permanente has just such a project: http://digitalrepository.aurorahealthcare.org/jpcrr/vol2/iss...