March 5, 2026
License Wars: Who owns the new ship?
AI and the Ship of Theseus
New code, same vibes—authors cry foul, devs cheer
TLDR: An AI-powered rewrite of chardet, done from tests to switch from LGPL to MIT, lit a fire over whether it’s truly new or a derivative. Comments split between moral outrage, calls to force all AI outputs under GPL, and memes—signaling a bigger fight over future software ownership
The internet went full courtroom drama over a maintainer using AI to rebuild the popular text‑detection tool “chardet” from just its tests, aiming to switch from a restrictive copyleft license (LGPL) to the more permissive MIT. Original author Mark Pilgrim says the remake is a “derived work.” The current maintainer, after 12 years on the project, insists it’s a fresh build: faster, multi‑core, different design, and validated as distinct by plagiarism checks. Cue the Ship of Theseus analogy: if you toss every plank and rebuild, is it the same ship, or a new one?
Commenters did not hold back. One furious voice compared the license fight to wanting a free car and snapped, “Are you like, 8 years old?” Another went nuclear: since AI was trained on open code, “any codebase that uses LLM code is now GPL,” which would instantly flip proprietary apps open. Others shrugged that the open‑source license zoo is collapsing to two choices: MIT or closed. The memes rolled in with “the slop of Theseus,” plus side‑eye about companies loving AI remakes—until their own frameworks get remade. Practical folks dropped alternatives like juniversalchardet. The vibe? Equal parts legal philosophy, soap opera, and gleeful chaos over who actually “owns” code when AI can rewrite it overnight
Key Points
- •The chardet library was reimplemented from scratch using only the API and test suite to enable relicensing from LGPL to MIT.
- •Mark Pilgrim, chardet’s original author, objects to the new version and considers it a derived work, while the maintainer views it as new work.
- •The new chardet implementation is described as faster, multicore-capable, and architecturally different, with distinctiveness reportedly validated by JPlag.
- •The author argues that low-cost AI-driven rewrites may weaken practical enforcement of copyleft licenses like the GPL.
- •The article highlights broader implications, including possible legal uncertainty over AI-generated code and renewed debates between copyleft and permissive licenses, with trademarks suggested as a control lever.