What happens when you press send to ChatGPT

What actually happens when you press send to ChatGPT
14 minutes by Alex Xu

This article explains what happens when you send a message to ChatGPT. Your text travels through secure channels and gets converted into tokens that the AI model processes. The system uses transformer technology to predict responses token by token, streaming answers back in real-time while applying safety checks and optional memory features.

A free course on real-time analytical data foundations
sponsored by Tinybird

Real-time isn’t a buzzword, it’s the performance edge for companies like Vercel, Canva, and Framer. Tinybird helps them transform massive streams of data into instant insights and real-time user experiences. To share what we’ve learned over the last 5 years running Petabyte-sized clusters, we built a free course on real-time analytical data foundations designed for developers who want to master the basics and build faster, smarter data products.

Programming with less than nothing
7 minutes by Josh Moody

Josh presents a funny story about a programmer implementing FizzBuzz using only combinatory logic. Instead of writing a simple loop, he builds the entire solution from just two basic functions in combinatory logic. The code becomes extremely complex and unreadable. Combinatory logic is Turing complete but impractical for real programming.

That time I trashed the company mainframe
5 minutes by Andrew Wulf

Andrew found a critical bug in an assembler by manually reading through the source code line by line. The bug caused endless loops that crashed an IBM mainframe. After studying the code and researching hash functions at the library, he identified the infinite recursion problem and reported it to the vendor. This experience taught him that reading and understanding code is just as important as writing it.

Build your own database
6 minutes by Nanda Syahrasyad

Nanda explains how to build a key-value database from scratch. It starts with a simple approach of storing data in files but shows this becomes slow for updates and searches. The solution involves using segments to prevent infinite file growth and indexes to speed up lookups. The final design uses LSM trees, which combine in-memory storage with sorted on-disk files. Major databases like DynamoDB use this approach and can handle millions of requests per second.

SQL anti-patterns you should avoid
4 minutes by Jordan Goodman

Jordan talks about some of the common and high impact SQL anti-patterns he has seen from experience that can make queries and pipelines difficult to maintain, or have slower than expected performance. These issues can compound, causing erosion in trust in data, and slower query development in general.

And the most popular article from the last issue was:

• Vibing a non-trivial ghostty feature