March 9, 2026
Fast code, furious comments
Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
AI rewrite drops the “share-your-code” rule; devs split between cheers and outrage
TLDR: A blazing-fast AI rewrite of chardet switched from a “share-your-code” license to a permissive one, sparking a brawl over law vs. ethics. Commenters split between celebrating speed and freedom, warning of a shrinking commons, and questioning whether the AI’s training and the rewrite are truly “clean”—a precedent-setting clash
An AI-assisted do-over just lit up dev land: the popular text tool chardet dropped a turbocharged Version 7.0—“48x faster” and multicore—after maintainer Dan Blanchard used Anthropic’s Claude to reimplement it from the public interface and tests, not the original code. Then came the plot twist: the license flipped from LGPL (a “share your changes” rule) to MIT (no such obligation). Original author Mark Pilgrim objected on GitHub, while open-source heavyweights Armin Ronacher and Salvatore “antirez” Sanfilippo said the move is lawful. The essay that kicked this off asks: does legal equal legitimate?
The comments? A popcorn factory. One camp shrugs that the outrage is performative—“just complaining”—and cheers the speed. Another camp says direction matters: GNU’s reimplementations freed users; this one removes the “give back” fence around the commons. A top meme compares the method to “throwing paint” and accidentally making Mickey Mouse—aka, legal fig leaf or real clean room? Others escalate to the source of it all: was the AI trained legally in the first place? Meanwhile, a gallows-humor thread coins a new category—“slop-licensed”—next to proprietary and free. It’s a full-on culture clash: pragmatists who want faster, freer-to-use code vs. idealists guarding the share-it-forward ethic. And nobody agrees where law stops and legitimacy starts.
Key Points
- •Dan Blanchard released chardet 7.0 as a ground-up reimplementation credited to Anthropic’s Claude, with major speed and multicore improvements.
- •The library’s license changed from LGPL to the permissive MIT license in the new release.
- •Blanchard says he used only the API and test suite with Claude, citing JPlag showing under 1.3% similarity to prior code and claiming an independent work not bound by LGPL.
- •Original author Mark Pilgrim objected, arguing LGPL requires derivative works to keep the license and that a maintainer’s exposure challenges clean-room claims.
- •Open-source figures Armin Ronacher and Salvatore Sanfilippo supported the relicensing’s legality, while the article argues legality does not equate to social legitimacy and that moving from copyleft to permissive reduces commons protections.