Your question @proc0 did make me think more about the value I've gotten from GH vs. what I've given back - I've signed up for the annual subscription plan as a result of that thought process.
33 karma · joined July 19, 2020
Currently solving problems at Remitly for immigrants - we’re hiring - reach out!
I wrote it because I'd directly seen or heard colleagues complain of real proposals to use git KPI as serious, primary perf measures for ICs. Linking to the satirical example of why poor measures result in poor outcomes can be a pretty effective tool for resolving the discussion to a good place ;).
This was a fun project to figure out and frame with the readme copy.
From the GH perspective I'm guessing they run some offline batch aggregation jobs to roll these stats based on the 24h timeframe. I capped the total commits to an arbitrary but polite # to avoid DoSing their stats pipeline after skimming ToS and seeing nothing that said - don't do this. It appears that they do monitor for excessive usage and care more about large objects than many commits: https://docs.github.com/en/github/site-policy/github-accepta...
I said something that would reasonably be interpreted way more broadly than what I meant, and was rightfully called out for it by a few folks.
My intent was to in a pithy way support how it's maybe not a great comparison to pull the top of one field and compare to the median of another when discussing compensation.
Bad take on my part in support.
Take the SCOTUS clerks from the last 10y. Pick a set of the same size of bureaucrats in one role from any time and place in history. Set them loose on a well translated set of each others' laws with a series of cases covering edges in the law to provide opinions on.
Who creates better rulings?
Do you think there is a set that does a better job at switcheroo2020? If not - does that mean that recent clerks meet the best at what they do criteria?
If you provide me the bureaucrats I'll: 1. Be super amused. 2. Happily go read about them and roll back my bad take.
The easiest target I'd go for is - current federal judges. Maybe that sinks me by itself here - in spirit I don't think so because the funnel overlaps heavily - and the theory is that the current funnel creates a historically rare result.
Dropped a line in above. I'm assuming a normal distribution of intelligence from a nature perspective that has not changed much in the last 2k years, and then on top of that an enormous nurture influence.
Right now there are a lot more people, with a lot more access to reading and writing education, and food - than there have been in the past. 90k legal graduates/y is a huge output. If you assume the filters that put folks into clerkship are a mix of meritocracy and then some elbow-rubbing noise - the end outcome should still be extremely talented compared to other groups of thinkers in the past. This does also imply that the same should be true for other current intellectual institutions that filter from large pools.
The assumption breaks down to the extent the filters aren't meritocratic. I'm not super qualified to speak to that - and there's at least one qualified opinion above expressing that the mix maybe leans more towards work and networking than ability - so maybe I'm wrong here.
Curious about your view on a qualified take.
By "some of the smartest individuals..." - I mean that there is a rare bar for work and aptitude here in the last ~3 decades if you look at a set of technocrats or thinkers historically.
Right now population * access to quality nutrition * access to early childhood and later education * interested in law school in the US produces a funnel of thinkers (90k graduates/y with ~19 years education) that is larger than almost any that has existed previously.
As some points of reference - some estimates put Rome at its peak at ~70mm citizens, the Tang Dynasty at ~80mm - both with access to fewer of the above conditions for the bulk of citizens. The Qing dynasty can hang on population in this thought experiment at 450mm, but the ratio of nutrition and education access was considerably lower.
If you compare SCOTUS clerks to actual luminaries like Gauss, they compare very unfavorably, but as a group I do feel that they are "some of the smartest individuals at what they do" that we've seen.
Is it possible that your high accomplishment in the field has given you an overly pessimistic view of the skillset required by your peers?
Your insight on the topic is super interesting - my background is as someone who was admitted to an elite law school and chose to follow their programming hobby instead - this part "However, even the absolutely best situated candidates don't have a more than average shot at success at their SCOTUS clerkship application." - makes me think that while there is a lot of networking involved here, the attainment bar is also - incredibly, unreasonably high - to be within striking distance, even if the last mile is networking, grind, and luck.
First paper does seem to substantiate article.
1. The lift is specifically strong for the judge they clerked for, and not for all of the judges 2. Federal clerks (also wildly competitive roles even if a bit less fancy) win about as much as scotus clerks from their analysis of another other data set, but don’t show some crazy single justice lift
Interesting stuff
Choosing to go all in on one is a personal choice that is going to be a poor fit in tons of cases.
The author’s product is based on folks accepting this premise and seeking tools to do better.
I do think the setup to get to the substance is a bit long and indirect - but it’s not a given, especially at series B/C startups that folks think about production monitoring as part of the QA process so I do like to see more pieces on that.