johnmyleswhite

People migrate -- they always bring their DNA with them and their language. No? The connections between these have been studied for a long time -- for example, this popular book was published 20 years ago: https://www.ucpress.edu/book/9780520228733/genes-peoples-and...

I still use Julia a lot and still think it is the best bet for a dynamic language with acceptable speed. I don't work on many OSS projects anymore, so I'm not an active contributor, but I'm not contributing to other languages in my spare time either. The projects you see in my "about" page are my paid efforts.

I think Julia's adoption is still not at all comparable to Python or R, but neither of those languages were comparably popular at this point in their own lifetimes. I think people underestimate how slow change is here.

Julia absolutely could fail to compete in the long run, but it is still growing and it is improving.

Given you're referring to me and my career, can you explain what you mean about "worked on large community Python projects"? I don't see that text on the page you're citing. Is it an interpretation of "I’ve been part of teams working on a variety of things, including extensions to the Python language and A/B testing"?

It seems more interesting to discuss the implied underlying situation here than focus on any individual actor's tone.

It seems from the interaction that a part of IBM has (1) taken software that explicitly has no warranties, (2) repackaged it and sold it for profit by unilaterally adding new warranties of their own creation, and (3) attempted to redirect the burden of compliance with those warranties to the original authors (who had explicitly disclaimed any such warranties).

Have you met any IC8+/D1+ working for Meta in the US?

Any chance this could be shared somewhere that's not Slideshare?

> Not a lot of people realize that Pandas was inspired by R, and in particular the Tidyverse model of handling rectangular data frames, created originally by Hadley Wickham.

Can you give a citation from this, preferably from Wes himself?

Does the article you linked not show that a loop is 10x slower than vectorization for computing square roots? The fact that 10x is better than the 60x slowdown for vapply isn't really evidence that loops are a reasonable alternative to vectorization yet.

Those inequalities often produce confidence intervals that are excessively conservative.

Julia's broadcasting syntax depends on having Julia-style compilation to be efficient. It's not clear to me Python could ever provide the same behavior.

This seems pretty off-base to me. A lot of folks with PhD's in physics are working in industry. And a lot of the folks working on distributed systems don't have PhD's in that field.

I imagine one factor is that there are lots of OSS projects that are simultaneously startups in the data space these days.

As others have said, I agree that you should seek professional help and find a way to re-evaluate your thinking about your career. From this post, it seems like you've gotten stuck in a feedback cycle of resentment. To make that more clear, note that many things you've said here would seem to imply you look down on your coworkers and/or on the work they do. There's the obvious example of "I was hard working and intelligent than him", but more telling are phrases like "I am extremely unhappy being an IC", "I have 25 years of industry experience and I feel I should mentor people instead of moving JIRA tickets around" and "Yet here I am being ‘advanced beginner’ in a different role every couple of years doing grunt IC work" suggest that you look down on the people who actually get things done at a company. How many IC's want a manager who views them as beneath them? What company would be exercising good judgment putting you into a role in which your resentment might trickle down to your reports?

I'd suggest taking a break and working with someone who can help you process your feelings and move past them.

The relational model's type system is contained inside of this work as the schema component.

I think this is a genuine possibility given that the poster describes themself as an L6, whereas a FB engineer would typically refer to themself as E6 or IC6.

If you work through the math for CUPED, you'll see that only the aggregate correlation term can be interpreted as being "per-unit" -- all of the other terms in the equation are the same terms you use in your post.

Insofar as I think your intuition is leading you somewhere, I think it's leading you towards a realization that a "diff in diff" approach rather than regression adjustment can increase variance in some settings. But regression adjustment is provably better in essentially all circumstances: the only settings in which it is ever worse than no adjustment are outlined clearly in https://projecteuclid.org/journals/annals-of-applied-statist...

Negative zero was introduced because of branch cuts: https://people.freebsd.org/~das/kahan86branch.pdf

This comment seems a bit odd given that the connections between physical neurons are qualitatively very dissimilar from the connections in artificial neural networks, so invoking the human brain seems like a red herring in discussing AI methods.

I believe NFLX doesn't hire the first two entry levels of engineers that other firms have.

Not the other way around? My experience is that the US uses 1-based indexing for floors in buildings and that Western Europe uses 0-based indexing for floors in buildings.

That one already got fired: https://www.nbcwashington.com/news/local/maryland-company-fi...

Isn't the reason that magrittr was able to use that syntax that it already existed, so your proposal is a breaking change?

What do you mean when you use the word "bytecode" in reference to Julia?

One of the important facts about OSS development is that the number of users doesn't scale linearly with the number of "free hands. There are less than 2 FTE's working on Python at any given time these days.

Isn't this a problem in any system that allows function names to be overloaded? Don't you have the same issues if (a) your system has single dispatch and classes that can be extended outside the module in which they're originally defined?

Yes, the nearest neighbor search problem is exactly the problem of finding nearest neighbors: https://en.wikipedia.org/wiki/Nearest_neighbor_search

I'm not familiar with DBSCAN as a NN search system. Do you have links to using it for that purpose?

I agree with your assertion that most users could get away with 10,000 bits. My point is only that there's always going to be somebody who still tries to do something that fails, so you have to pick some set of tradeoffs. Yours seems reasonable.