Singal has claimed multiple times over the years that his mistaken interpretaion actually means his conclusions are *stronger*. But I cannot find any methodology by which he can justify that claim. Instead, he begs the question and commits handwaving.
Moreover, Singal has failed to address commentary by the very author of the study he cited which directly contradicts the erroneous conclusions he draws.
Jesse already admitted that he made a mistake in lumping "lost to follow-up" participants in with the category of "desisters". And while he has updated at least one piece that relied heavily on this analysis, he did not retract the conclusions.
He should have, however. Because Singal is publishing a book on the reproducability crisis in psychology, so you would think that he would be able to sight-identify a glaring reproduceability error in a paper.
Whoever the product owner is behind Virginia's vaccine registration system comms—you're doing great, because this is IMO an example of really good communication.
I'm gonna pour a coffee and explain why.
First—this very message is simply a reminder that you've already registered. Registering for things is, for the user, a very non-sticky experience. It's easy to forget you registered for something, even if (especially if) it's important. Regular gentle reminders are good.
This helps the user, but it also helps data quality and preventing accidental system abuse.
It's clear what this message is: it's not a call to action. It's just a gentle reminder of what's necessary.
I’m reading @Okwonga’s novel and I can’t say I am devouring it. I am savoring it, page by page, sentence by sentence. It is simply the most relatable thing I have ever experienced in print.
The way he tells of the ways one finds themselves in a relationship with Berlin, Berlin the experience, Berlin the city of perpetual adolescence, it is something you cannot understand until you have experienced it for yourself.
People talk about New York because New York wants you to talk about New York.
Berlin is a shrugged shoulder, and whether that shrug represents indifference or acceptance or devastation is something you’ll never quite figure out.
There's a shocking amount of re-refined syntactic sugar floating around the python ecosystem these days that I don't think a lot of devs know draws directly from many layers of trying to fix FORTRAN, a perfectly fine language.
MATLAB sits in the family tree of way more things than you might realize.
pandas was basically like, "numpy should be more R-like" and numpy was basically like, "python should be more MATLAB like" and MATLAB was like, "what if the notion of developer ergonomics were to be invented in like 30 years"
Single Source of Truth is basically an impossibility, not because of technical reasons per se, but because “truth” is a notoriously hard problem and it presumes that your technology systems are designed to cover the full space of analog realities.
It’s a losing battle from the start to try to architect a system that is structured around single sources of truth. The problem becomes more complex with every transformation and rapidly intractable.
It’s much better to have a notion of automated data quality.
Data pipelines are messy because they move data that reflects real-world processes, which are messy.
So I’m in the process of migrating some of my stuff off of Google Cloud, and I’m committing to doing this right by going full IaC and learning terraform and Azure at the same time and... this is easier than I thought it would be at first? That’s kinda cool.
I’m migrating off of @googlecloud because I am really unhappy with the way Google treated @timnitGebru and the way they handled this is indicative of what I consider to be a deeply mendacious way to wield ethics in technology.