Balaji Srinivasan

Network effects, yes like how BTC is owned by a tiny pool of oligarchs who need other users to transfer real capital in for their zero sum pyramid scheme.

One important point: if we actually include all 7 billion
people on the earth, most of whom have zero BTC or
Ethereum, the Gini coefficient is essentially 0.99+. And
if we just include all balances, we include many dust
balances which would again put the Gini coefficient at
0.99+. Thus, we need some kind of threshold here. The
imperfect threshold we picked was the Gini coefficient
among accounts with ≥185 BTC per address, and ≥2477 ETH
per address. So this is the distribution of ownership
among the Bitcoin and Ethereum rich with $500k as of July
2017.

In what kind of situation would a thresholded metric like
this be interesting? Perhaps in a scenario similar to the
ongoing IRS Coinbase issue, where the IRS is seeking
information on all holders with balances >$20,000.
Conceptualized in terms of an attack, a high Gini
coefficient would mean that a government would only need
to round up a few large holders in order to acquire a
large percentage of outstanding cryptocurrency — and with
it the ability to tank the price.

With that said, two points. First, while one would not
want a Gini coefficient of exactly 1.0 for BTC or ETH (as
then only one person would have all of the digital
currency, and no one would have an incentive to help boost
the network),…

That preprint [1] is a great example of how far ahead of the skis you can get with motivated reasoning, bad math, and no peer review.

They recruited people over Facebook at a time that California's COVID tests were backlogged (with ads not presented in the paper, with links to signup that could, anecdotally, be shared out); in that situation, avoiding selection bias towards people that were sick with something previously and want to know about potential immunity would be rather tough. (Per WTO protocol, they asked about symptoms in their sample but poof for some reason that data isn't in the paper).

They got a 1.5% positive rate in their sample with 50 individuals testing positive. They then immediately applied demographic weighing to move that raw number to around 3%. Only then did they contemplate the false positive rate, taking the data of the manufacturer and 30 samples as a point estimate to figure out a 95% CI of positives around that 3%.

That's a no-no for correctly propagating uncertainty. The statistically legitimate way to do that would've been to generate the 95% confidence interval for belief about the test's false positive rate and carry that forward in the demographic reweighing. That basic math would've avoided the frankly laughable result that a randomly selected person in Santa Clara county could have a 5% chance of having had COVID-19 while in the same county people seeing a doctor and getting tests through their doctor are only around 10% positive for…

I followed these guys pretty closely back then and this is close to my recollection as well.

One addition I had to check on, which confused me at the tilt was how Balaji was so closely involved in 21 yet also a partner at Andreesen Horowitz, which I had read about their daily and weekly routine and was clearly a full time job.

Balaji went from founder to "advisor" of 21 pretty quickly, becoming a partner of a16z in the interim. After things went to shit at 21 he went back as full time CEO. Reasons for how that went down are speculative.
The results are not though. The exit is real but wow is it questionable. Conflicts of interest seemingly all over the place. I don't know who else is on Coinbase's cap table but I'd love to hear their takes on the acquisition.

You can read Balaji's congratulatory take in his own words on what he did here. [0]

I also recall Balaji positioning himself for a position in the Trump administration which saw him scrub his entire twitter history. If anyone recalls more details from that episode feel free to add them.

[0] https://medium.com/@balajis/the-turnaround-2d145589d814

You don't get a gourmet meal by starting with garbage and reprocessing it and adding garnishings.

Most notably

"But there’s continued concern about the methodology. Volunteers for the test were recruited from Facebook — so its sampling was neither random nor representative of the community as a whole."

For an expert review of all the problems please see

https://medium.com/@balajis/peer-review-of-covid-19-antibody-seroprevalence-in-santa-clara-county-california-1f6382258c25

In the midst of a crisis the poor quality work of the scientists suggests that they either don't take the matter sufficiently seriously or don't possess the necessary skills to deserve our trust. It's a dangerous time to take bad advice especially at a time people are so desperate for good news.

Sure!

Andrew Gelman (Stats at Columbia) had a commonly shared piece: https://statmodeling.stat.columbia.edu/2020/04/19/fatal-flaws-in-stanford-study-of-coronavirus-prevalence/

Also a good dive into the issues: https://medium.com/@balajis/peer-review-of-covid-19-antibody-seroprevalence-in-santa-clara-county-california-1f6382258c25

Mercury News also had a good article covering a lot of this: https://www.mercurynews.com/2020/04/20/feud-over-stanford-coronavirus-study-the-authors-owe-us-all-an-apology/

And yes, lots of twitter discussions from folks in the field, e.g. Natalie Dean of University of Florida https://twitter.com/nataliexdean/status/1251309217215942656
and Trevor Bedford (Fred Hutchinson) https://twitter.com/trvrb/status/1251332447691628545 and others.

Sampling uncertainties aside, this is super sensitive to mis-estimating the false positive rate. Which they seem fairly likely to have done, using an estimated test specificity of 100% (with error bars) in one branch of their analysis after getting 30/30 negatives against pre-COVID serum samples. Too small a sample to get that confident in an atypically high specificity, especially if that collection of samples was disproportionately low antibody (e.g. summer blood rather than post cold/flu season blood).

That's to my mind the most consequential misestimation of the test characteristics, but Balaji Srinivasan details more:

https://medium.com/@balajis/peer-review-of-covid-19-antibody-seroprevalence-in-santa-clara-county-california-1f6382258c25