
Python Bytes #473 A clean room rewrite?
Mar 16, 2026
A heated licensing clash over an AI-assisted clean-room rewrite of a popular encoding library. A look at Refined GitHub and small UI boosts that speed developer workflows. A deep dive into PgDog, a Postgres proxy that handles pooling, load balancing, health checks and sharding. Practical patterns for building and testing agentic systems and cautions about unreviewed AI-generated code.
AI Snips
Chapters
Transcript
Episode notes
Fair Use For AI Training Remains Unclear But Trending Transformative
- Emerging judicial consensus may treat training general-purpose models on copyrighted material as transformative fair use, but rulings vary by case.
- The U.S. Copyright Office still warns using copyrighted materials for model training may be prima facie infringement.
Resolve License Drama By Releasing v7 As A New Project
- Michael suggests a pragmatic fix: treat chardet v7 as a new project name (e.g., char.2) and keep v6 stable under its original license to remove legal friction.
- That preserves the faster, multi-threaded v7 while honoring users who depend on the older LGPL release.
Improve GitHub Productivity With Refined GitHub
- Try Refined GitHub to subtly improve developer workflows with many small UI upgrades like CI icons, PR workflow links, tab support in comments, and highlighting useful issue comments.
- Brian installed it and notes it adds 100+ unobtrusive improvements and has ~30,000 stars, making it trustworthy for heavy GitHub users.

Also pretty good and related: