loading...

Friday Blast #49

horia141 profile image Horia Coman Originally published at horia141.com on ・2 min read

What do you believe now that you didn’t five years ago (2018) - interesting read about centralized vs decentralized systems. And how five years ago people believed they’d take over (I mean, more than in the case of email, the WWW etc), and that hasn’t really happened. And it’s probably never going to the way proponents would want.

For mathematicians, = does not mean equality (2018) - it means what they want it to mean. Much more so than programming languages, mathematics depends a lot on the context of readers. While things could be done super-formally, nobody does it. This article explains why, and how the simplest of mathematical notations carry a lot of baggage.

Operationalizing Node.js for Server Side Rendering (2018) - well, if this won’t put you off SSR. An interesting article from AirBnB on how they slowly migrated from a Rails “regular rendering setup”, to a SSR frankenstein to a simpler SSR. In all cases, the fact that we’re dealing with a compute-intensive operation changes the calculus of building such services. From how you allocate machines, to what happens when there’s more requests than capacity for them. In backend land, we don’t give enough credit to how puny the computation we do is relative to the I/O, and how easy that makes our life.

Circular law for random matrices (2018) - applied math warning. The eigenvalues of random matrices (ie matrices with components drawn at random from some distribution), neatly fill the complex unit circle under mild assumptions.

Scaling consistency (2018) - how to build “cross-sectional” review boards, architecture committees etc in growing organisations so that they’ll have the maximum impact and least repercussions. Stuff you can’t read in books.

Learning market dynamics for optimal pricing (2018) - a neat read about modelling problems. This is about predicting the price rooms in AirBnB will be rented at. There’s a lot to unpack here, but the model is equal parts business knowledge, applied machine learning and old fashioned time series forecasting. I really dig it when I see systems like this being advertised. Sure, Deep Learning is all the rage, but a lot of stuff happening is mostly “the simplest model that’ll work”. In many cases that’s a classical linear mode, or some tree etc. And in this case the powerful tools from classical statistics provided the juice to get the thing in a good state. Be on the lookout for some Fourier series as well.

Posted on by:

horia141 profile

Horia Coman

@horia141

Leading the Bolt teams in Bucharest. We're working on cool products like food delivery, and interesting and challenging platforms like geo, A/B testing, user accounts and route tracking.

Discussion

markdown guide