Show more
22 boosted

When “buy a diversified low-cost index fund” became dogma that my college roommates and PhD-holding coworkers alike intoned to signal that they are sophisticated adepts in the cult of Vanguard who have unlocked the secret to successful retirement investing, years and years ago, my contrarian hackles stood.

I couldn’t offer a nice alternative, and neither does Burry (other than “invest in my hedge fund”?), so nobody cares what I think—just, holes in the dogma are easy to find if you study & dig.

‘passive has removed price discovery from the equity markets. The simple theses and the models that get people into sectors, factors, indexes, or ETFs and mutual funds mimicking those strategies—these do not require the security-level analysis that is required for true price discovery.’

bloomberg.com/news/articles/20

HN has sent me a couple of very high quality stories recently! I like Michael Burry a lot, this interview is fascinating! 🤫 I’ve avoided index funds for years.

22 boosted

Airports use up a lot of land. I shouldn't be surprised but I am. Also, nice location choices for golf and large family landowners.

github.com/fasiha/nextprod-py

Publishing package to PyPI is not in muscle memory (as publishing to npm is) so I'll do it later.

The person who ported Ebisu to Go did this, a blind port from JS to Go. If I hadn't seen that I'd have wasted untold hours figuring out the algorithm Julia nextprod is using and rewriting it Pythonically. Direct syntax translation FTW.

Moving on with my life.

(This is for . I want to ship that library. It's been four years now, but very nearly done.)

If you found a problem, then ameliorated it with code, even if it's shit jQuery or ObjC soup, that's meaningful!

If you struggled to maintain your understanding of what was going on, or to keep adding features, that's totally fine!

If it made you want to get a bit better at the craft of coding so you can solve more problems a bit faster, come see me!

prog21.dadgum.com/87.html be the skull-blood kid!

Being able and willing to think deeply about someone's needs, and to iterate on your understanding by getting criticized by them, is I think a crucial skill in programming.

Being able to defend your choices—what feature they enabled or what obstacle they overcame. To describe blind alleys and failed experiments. That's what I look for.

We can teach you static types, pointer arithmetic, CPU caches, matrix decomposition algorithms. But we can't make you give enough shits to ship something.

“You can passively consume hundreds of articles and podcasts and learn far less than shipping one side project a year…” mobile.twitter.com/r00k/status

I have more to say about this too. Replace "articles…podcasts" above with "computer science courses".

I want to see shipped projects, that you made for someone (yourself, your grandma, whatever) after carefully deciding why using something existing wouldn't work. That they criticized.

Not projects your professor or your intern manager gave you.

People have been writing since before writing (Gilgamesh, Plato, etc.) about how knowing something is very weakly correlated with changing behavior.

Attenuating the constant low-grade background radiation of anxiety in your life won't happen by knowing something, like about trade wars or occupied territories or etc. It happens by shifting your perspective on your reactions to the events in your life, by changing habits.

And certainly don't argue based on what you read in the news.

‘Reading about the trade war with China doesn’t make you smarter—it gives you something to say at dinner parties. It gives you the illusion that you have the vaguest idea what is happening in our enormously complex world.’ tjcx.me/posts/consumption-dist

I have more to say about this: if you can't resist news by realizing it's giving you (lethal) illusion of understanding, try this—assume you had a demon that perfectly explained the trade war to you. Would it alleviate your or anyone else's suffering?

I'd like to see that platform be able to support placing bets on any published (VaR) feed.

"You think you can break my system of predicting center mass of Wikipedia editors? Bring it." Then it's a short step to "You think you can break my system of tail-bounding S&P500 PNL? Bring it."

According to , taking bets on your VaR system from all comers is a crucial piece of strengthening it and of quantifying risk. If 5x people today bet it'll break today versus yesterday… !!!

The plan for is to not only

- get familiar with mechanisms of predicting (including the major subtask of feature engineering—imagine all the different data streams that might help you predict how many people are gonna show up to the French language to edit it in a given 24 hour period), but also

- make a platform that accepts others' bets on whether a VaR break will occur. Even if this is with toy money, working through the tech hurdles there is very exciting.

22 boosted

The world needs more signs like this. ^_^

Curiously, despite the poetic meme being very Chinese, the Chinese on the sign may itself be a bad translation from the English:

languagelog.ldc.upenn.edu/nll/

OH: “we stayed at a town so small they didn’t have Uber”.

The horses 🐎 model appears to by multinomial logit, bread and butter for statisticians. The hard part is the feature engineering. The really hard part is (I’m told) finding someone to take your bets.

I want to find something Bayesian first, that produces probabilistic predictions, so I can produce figures daily.

‘Benter’s model… monitored only about 20 inputs—just a fraction of the infinite factors that influence a horse’s performance, from wind speed to what it ate for breakfast. In pursuit of mathematical perfection, he became convinced that horses raced differently according to temperature… he entered the data into his computers—and found it had no effect whatsoever on race outcomes. Such was the scientific process.’ bloomberg.com/news/features/20

Doing this at , predicting Wikipedia editorship.

"Like the empire of Darius I, the realm of the Arsacids covered territories in which many languages were spoken. In Iran these were mainly Middle Persian, Parthian, Sogdian, Choresmian and Bactrian; further west they spoke Armenian, various Caucasian languages and Babylonian; in Mesopotamia and other parts of the empire, the language was Aramaic in its different variants, and in the Greek poleis such as Susa and Seleucia-on-the-Tigris it was Greek." —Josef Wiesehöfer, "Ancient "

😱!

Show more
Octodon

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!