Backpressure for dummies
A toy example showing how lacking backpressure may lead to failures and how to add it.
Interesting readings of the month
Great summary from Noam Bardin, Waze CEO, of his seven years at Google, following Waze acquisition, and how this change affected how Waze operated. Although probably none of the points he makes come as unexpected, hearing them from someone who went through that experience adds a lot of value to them. Also, I always find it refreshing when an executive is able to preserve such a level of candor when sharing his opinion in the open. The shift in mindset from “working on a mission” to “working on your career progression” is a particularly strong point, which I think to an extent made me prefer jobs at start-ups so far. I also particularly appreciated his point on how a successful employee at one stage of the company may not be so at another one, and it may be necessary (and possibly good for him) to fire him.
Really nice inside look into how being low income may turn what an average income person considers background noise into daily concerns. Being low income makes it almost impossible to go beyond covering monthly expenses. One consequence of this is the struggle to being able to purchase goods less prone to failure or likely to generate negative outcomes (a car that doesn’t break down, a house in a safer neighborhood). Another consequence is how harder it is creating a savings buffer that could shield you from unexpected negative events likely to make you really broke. This article is valuable not only for the practical examples in the different life dimensions (house, healthcare, transportation) of what it means to be poor-sh but also for giving non-poor-sh people a different reference point when evaluating their own hardest times. Everyone’s judgment of their own or others’ situation is very much based on their current or past experiences. Covid may have been by far the lowest point, financially and professionally speaking, for many white-collar workers who like me entered the western world job market after the 2008 financial crisis. At the same time, although we may feel miserable for a missed promotion, finding ourselves unemployed or furlough, not many of us ended up broke or without a roof over our head because of this. Still, these negative events may have had a large emotional toll on our well-being. By considering a different perspective on what a “bad” financial situation is it should become easier to weigh your own in a more balanced way as well as develop some empathy toward others less lucky than you.
For some time I have been trying to get a better understanding of Monte Carlo methods. I started from the relative chapter in All of Statistics, which although quite dry, gives a decent initial high-level view. Then the Tweag blog series on the topic gave me a better practical understanding of the how. Finally, Betancourt’s Conceptual Introduction to Hamiltonian Monte Carlo has been a game-changer in terms of understanding Hamiltonian MC. This method is possibly the most commonly used one as well as probably the toughest one to understand (in particular, for someone who is not from a physics background). This paper is among the best ones I have ever read for what concerns clarity in communicating a complex topic. The work explains how when moving to a higher dimensional space a random walk exploration of the parameter space becomes so inefficient to be unfeasible. In presenting the source of the inefficiency it also sets the ground for what it takes to get an efficient exploration, ie focusing the exploration on the typical set (ie the location in the parameter space where both the volume and the density are large enough). Finally, it introduces the Hamiltonian process and explains how this is able to constrain the exploration to the typical set.
Time taken to write post: 2 hours
A toy example showing how lacking backpressure may lead to failures and how to add it.
A simple piece on how to patch objects and use pytest fixtures in tests at the same time
Using various tools to make python environments working automagically also with emacs
On the benefits of documenting your work and how it can impact your share of the pie
Looking back at 2021 to look forward at 2022
Several ideas worth remembering from Sandi Metz’s book 99 bottles of OOP
Can we do better than a straightforward pickle.dump?
Why interviewing time to time is useful and what I learned so far by doing so
Interesting readings of the month
Interesting readings of the month
Use case of the Strategy pattern in a data science application
Interesting readings of the month
Considerations on how practicing writing can help improving how we communicate and how we think
an overview of what eigenvectors and eigenvalues are and why they make PCA work
A possible pattern for replicating programmatically AJAX POST requests when scraping webpages using Scrapy