200 articles

Introduction

The previous article, ChatGPT's critique of OpenAI CEO Sam Altman's vision of the AI future was the 200th post on Bytepawn! I crossed the 100 article mark almost exactly 3 years ago in 2021 October, so it took 3 years to write 100 more. Below is a list of some notable articles from this second batch.

Favorites

1. The article I'm most proud of in recent times is Breaking Bell's Inequality with Monte Carlo Simulations in Python. As a physicist I learned about Bell's inequality at University, and have come across it multiple times since then, but it's never been intuitive to me what the real meaning behind it is. One of the themes of my intellectual journey over the years — well documented here on Bytepawn — has been the idea of using Monte Carlo simulations to better understand statistical concept. This article fits beautifully in this thread, as I realized I can formulate the Bell experiments as a game modeled in Python, and then play around with different systems to get a feel for the real meaning of the inequality. This article was also positively received on Hacker News.
   
   

2. I have continued writing many articles on A/B testing — this topic is definitely the most discussed here on Bytepawn. My favorite, and most useful article I've written is Five ways to reduce variance in A/B testing, which is the culmination of many previous articles on this topic, such as Reducing variance in A/B testing with CUPED. Another notable post on the A/B testing thread is Paper review: A Comparison of Approaches to Advertising Measurement, which explains in plain language the main results of the paper by Gordon and Zettelmeyer, that observational methods severely misestimate experiment lifts.
   
   

3. Continuing on the previous thread, in 2022 January I used a Monte Carlo simulation to beat the then-popular word game Wordle, which I documented in the 3 part series: Beating Wordle with Monte Carlo methods.
   
   

4. A cool feature I added to Bytepawn is Other Articles You May Like, the little blue box at the bottom of every article. I wrote my own custom Python implementation for this using Doc2Vec, which I explained in a 3 part series: Similar posts recommendation with Doc2Vec.
   
   

5. In the last couple of years I decided to rekindle my love for C++ programming. One of my hobby threads on this is writing toy implementations of an async in-memory message queue in various programming languages, one of them being C++. But my favorite recent C++ project was re-writing the standard wc Unix utility in C++ and in the process beating the standard implementation, which I documented in the 3 part series: Wordcount - Beating the system wc.

6. I've always been interested in Blockchain, as the underlying distributed and cryptographic primitives are fascinating to study. However, I've always been and remain highly skeptical of real-world applications, such as the many crypto coins, and later NFTs. In 2023 January I wrote a 5 article series on why I'm skeptical about Crypto and NFTs.
   
   

7. I've written many articles on leadership topics. By far my favorite topic is that of culture, and capturing it in Culture Docs, a sub-topic about which I've written 2 articles: first, Culture Docs: Facebook, Netflix and Valve, and then when we created our own Data Science Culture Doc. The most recent version of this you can see at maf-datascience.com/culture. Another interesting article is my recent How I manage my 1v1s.
   
   

8. On a very personal note, I wrote Why I don't live in Hungary, which lays out the many reasons I decided to move from Budapest to London in 2016 and then to Dubai 2 years later, and why I don't currently intend to move back home with my family.
   
   

Conclusion

Doing the research and programming for my Bytepawn articles has been incredibly rewarding both intellectually and professionally. Writing this blog is one of the best things in my life, along with spending time with my family and doing triathlons.

One recent change I made is to remove Google Analytics from Bytepawn. I used to check my viewership stats almost daily, but at some point I realized this was a pointless vanity exercise. I am in the lucky position that I don't need to monetize this blog, I write the articles for my own enjoyment, so I don't really care who or how many people see it. I still sometimes post an article that I thought came out particularly well to Hacker News, but I don't really care whether it takes off.

Since some of the articles I write here have become quite technical, I now view Bytepawn as a feeding funnel for writing rough drafts of ideas that could end up as Arxiv papers, like my recent Monte Carlo Experiments of Network Effects in Randomized Controlled Trials. I now have a personal goal of converting 1-2 Bytepawn articles to high-quality Arxiv papers annually.

Onward to the next 💯!