Matt Levine on the SEC case against App Annie
Your hedge fund is in the business of getting “edge.”
It wants to know things that other people do not know, in order to buy the stocks that will go up.
Another way to get edge, quite popular these days, is to buy “alternative data.”
The car-counting is a classic story of alternative data,
but the way alternative data mostly works in practice is that a lot of people use apps on their mobile phones, and
everyone involved in the mobile app business is harvesting data and frantically selling it to each other and to hedge funds.
[The collection and sale of app usage data as alternative data is fine if] everybody has consented to the use of their information, so you can use it freely.
If at any point in that chain — which can have more or fewer links than I wrote in the previous paragraph — someone doesn’t give permission,
then the data is misappropriated.
If you use that data in your trading, you are insider trading, trading based on misappropriated material nonpublic information.
Also the person doing the misappropriating — the person passing along data without permission — is also engaged in insider trading, if they know that you’re going to be using it to trade.
Yesterday the U.S. Securities and Exchange Commission brought a pretty important enforcement action against a company called App Annie Inc. and its former chief executive officer, Bertrand Schmitt;
Schmitt and App Annie settled for about $10 million.
App Annie got data from apps and sold it to hedge funds
without using best practices of fully informed consent
Basically if you are a company with an app, you can use an App Annie product for free to, like, see how good your app is
App Annie provides a free analytics product called “Connect” to companies that offer apps,
including public companies, which enables those companies to track how their apps are performing.
As a condition of their use of Connect, those companies provide App Annie with their app store credentials
to allow App Annie to collect their confidential app performance metrics (“Connect Data”).
App Annie tells Connect users that
it will generate estimates of app performance using their Connect Data,
but that it will only use Connect Data in aggregated and anonymized form to generate those estimates.
App Annie’s business model relied on selling estimates of how apps belonging to certain companies were performing.
App Annie sold these estimates through a paid subscription product called “Intelligence,”
these “estimates” worked like this:
App Annie used aggregated anonymous data from the apps it tracks to build a statistical model of how apps with certain characteristics perform.
It used that model to generate an estimate of how each particular company’s app performed.
Then it checked that estimate against how the app actually performed, which it knew, from collecting that app’s data, and corrected the estimate to reflect the actual performance.
To make these manual alterations, the Delivery Team looked at confidential Connect Data, including public company app performance data.
App Annie created a manual process
whereby a “Delivery Team,” which consisted of a subset of App Annie engineers based in Beijing, China,
made manual alterations to estimates
generated by the statistical model
before they were delivered to Intelligence subscribers.
if you are a hedge fund lawyer, now you are on notice that you should be asking your alternative data providers some tougher questions.
The American Style of Quotation Mark Punctuation Makes No Sense
There are different ways of combining quotation and punctuation marks.
In the American style, you almost always put periods and commas inside the quotation marks:
In the British style, you put periods and commas outside the quotation marks,
unless they are part of a complete sentence that is fully contained between the quotation marks:
Since a period marks the end of a sentence, it should not be placed before marking the end of the quotation
You can compare this with
nested or hierarchical structures, or with
stacks, or even with
first in, first out methods of computing, systems theory or asset management.
Under any comparison, the British style will seem preferable to the American.
You resolve the nested item first, before resolving the parent.
[The only exception is when] what is being quoted is a full sentence.
[where] We are placing the period to mark the end of the inner, quoted sentence.
Chapter 5. Hacker Writing Style
Hackers tend to use quotes as balanced delimiters like parentheses, much to the dismay of American editors.
It is counter-intuitive to hackers to mutilate literal strings with characters that don't belong in them.
Given the sorts of examples that can come up in discussions of programming, American-style quoting can even be grossly misleading.
When communicating command lines or small pieces of code, extra characters can be a real pain in the neck.
Apple CPU Gains Grind to a Halt and the Future Looks Dim as the Impact From the CPU Engineer Exodus to Nuvia and Rivos Starts to Bleed In
Apple in general was very clammy about the A15 comparison in the new iPhone reveal.
Instead of comparing it to the previous generation like they usually do, they opted to compare to ambiguous “competitors.”
The CPU is claimed to be 50% faster than the competition
while GPU is claimed to be 30% or 50% faster
depending on whether it is 4 cores or 5 cores.
The most important thing to note is that the CPU gains are identical from the A12 to A14 as they are from A12 to A15.
The GPU gains are quite impressive with a calculated 38.5% improvement.
This is larger than the A13 and A14 improvements combined.
They are sticking with a 16 core NPU which is now at 15.8 TOPs vs 11 TOPs for the A14.
the system cache has doubled to 32MB.
[We] also believes Apple moved to LPDDR5 from LPDDR4X
They state the new iPad mini has a 40% faster CPU and 80% faster GPU versus the previous iPad mini which contained the A12.
These are performance gains are generally paltry despite a huge increase from 11.8B transistors to 15B.
Furthermore, with next year’s A16 on the N4 process rather than N3, gains look to continue to slow.
Apple is clearly investing their transistor budget in the non-CPU aspects of the SOC.
Fixed function and heterogenous compute reign supreme.
The brain drain continues and impacts will be more apparent as time moves on.
As Apple once drained resources out of Intel and others through the industry, the reverse seems to be happening now.
In 2019, Nuvia was founded and later acquired by Qualcomm for $1.4B.
Apple’s Chief CPU Architect, Gerard Williams, as well as over a 100 other Apple engineers left to join this firm.
Rivos Inc, a new high performance RISC V startup which includes many senior Apple engineers.
We believe Apple had to delay the next generation CPU core due to all the personnel turnover Apple has been experiencing.
Instead of a new CPU core, they are using a modified version of last year’s core.
Regardless of the paltry CPU gains and potential core architecture delays, Apple is still the leader in performance per watt.
With Intel design teams starting to get back on track,
AMD executing almost flawlessly, and
Qualcomm coming in soon like a hammer with Nuvia cores,
we aren’t sure if this lead will be sustained.
The A11 to A12 generation was seen as Apple starting to asymptote out on gains with only a 15% gain, and
the A13 to A14 looked even more weak with 8.3% gains, but now with no CPU gains,
let's cross our fingers and hope the A16 brings a large architectural change.
Apple Watch Series 7 Tidbits
As noted by Steve Troughton-Smith, the CPU in the S7 chip carries the same t8301 identifier as the CPU in the previous S6 chip, which explains why Apple's performance claims are unchanged relative to the S6.
With the Apple Watch Series 5, Apple included the S5 chip that carried the same CPU as its predecessor but with the addition of a gyroscope
With the larger display on the Series 7, the S7 chip is likely to have been tweaked to continue providing customers with all-day battery life while still powering the larger, more advanced display.
Designing Beautiful Shadows in CSS
Creating a cohesive environment by coordinating our shadows.
Each element on the page should be lit from the same global light source.
The
box-shadowproperty represents the light source's position using horizontal and vertical offsets.To ensure consistency, each shadow should use the same ratio between these two numbers.
As an element gets closer to the user, the offset should increase, the blur radius should increase, and the shadow's opacity should decrease.
You can skip some of these calculations by using our intuition.
Using layering to create more-realistic shadows.
Tweaking the colors to prevent “washed-out” gray shadows.