A Byte of Coding Issue 395

Hey-yo,

It’s been a while since I’ve been so excited to work on the newsletter website. I’ve restricted myself to only working on it on the weekends, otherwise I’d neglect other more mundane work that needs to be done. I’m genuinely looking forward to setting up a good search feature that lets me search through previous issues AND the articles that were featured in them. I think the passion has been reignited by a combination of AI enabled programming and growing familiarity with using caprover to deploy apps. A well oiled deploying process is a huge boon, especially for someone who loves seeing their work out in the wild (let’s be realistic here though, there are only about 25 people on the A Byte of Coding website every day, so it’s about as wild as a small local park).

Also I’m pretty sure we’re not too far from another round of AI wrapper apps being launched as soon as Microsoft / Apple / any other OS updates their software to work with local AI models. Let’s see what the law of large numbers has to say about when we can expect that.

Anyway, here’s the issue.

Made possible through generous sponsorship by:

Avoiding Seemingly-Necessary Retain Cycles

Published: 27 April 2024

Tags: memory, compilers

Myles C. Maxfield explores methods to avoid retain cycles when implementing APIs that appear to require them, sharing insights from their experience with the WebKit team.

Some highlights:

• detailed explanation of avoidable retain cycles in programming, focusing on reflection objects within compilers

• presents a technique where all objects in a cycle share a single reference count, thus avoiding individual retain counts and simplify memory management

• solution includes practical steps like modifying retain and release methods and ensuring proper object destruction to manage lifecycle efficiently

Tokens for LLMs: Byte Pair Encoding in Go

Published: 25 April 2024

Tags: go, machine learning, webassembly

Eli Bendersky presents a detailed guide on implementing Byte Pair Encoding (BPE) for tokenization in Go, aimed at creating efficient token systems for language models.

Some highlights:

• BPE is a technique initially designed for data compression but adapted for tokenization to improve machine learning model efficiency

• includes a complete, albeit not speed-optimized, implementation in Go that can be used as a playground for experimentation

• provides insights into training the tokenizer, detailing the process of merging tokens based on frequency to build a vocabulary optimized for machine learning applications

The Developer's Guide to User Management

Published: 25 March 2024

Tags: sponsored, auth

WorkOS published an extensive guide on building modern day user management, diving into various auth topics like SSO, MFA, and even user provisioning.

Some highlights:

• Session management is a foundational piece of user management but is complex to build. Devs need to consider cookie-based vs. JWT-based auth and how to persist login state information for users that are part of multiple orgs

• At the 201 level, things like bot detection and org auth policies become more relevant

• 301 topics include invitation flows, email verification, and more

Reexamining FizzBuzz Step by Step – and allowing for more varied rules

Published: 28 November 2023

Tags: ruby

Tobias Pfeiffer reexamines the FizzBuzz coding challenge, introducing the idea of variable rules to enhance its complexity and adaptability in coding interviews.

Some highlights:

• discusses adapting FizzBuzz to allow for a dynamic set of rules, which makes the challenge more relevant and challenging for modern coding practices

• proposes and implements an improved version of FizzBuzz in Ruby, focusing on modularity and testability, which supports easy changes to rules

• post includes thoughts on potential extensions to the challenge, such as integrating new rules and varying output methods to suit different conditions and requirements