AI Reimplementation of chardet Library Raises Copyleft Licensing Questions

What Happened with chardet
Dan Blanchard, maintainer of the chardet Python library used by roughly 130 million projects monthly, released version 7.0 with significant changes. The new version is 48 times faster than its predecessor, supports multiple cores, and was redesigned from the ground up. Anthropic's Claude is listed as a contributor to this release.
The license changed from LGPL to MIT. Blanchard's approach involved feeding only the API and test suite to Claude, asking it to reimplement the library from scratch without directly looking at existing source code. JPlag analysis shows the resulting code shares less than 1.3% similarity with any prior version.
The Licensing Dispute
Mark Pilgrim, the library's original author, objected via GitHub issue, arguing that the LGPL requires modifications to be distributed under the same license. Pilgrim contends that a reimplementation produced with ample exposure to the original codebase cannot pass as a clean-room effort.
Two prominent open source figures weighed in with opposing views. Armin Ronacher, creator of Flask, welcomed the relicensing, stating he considers "the GPL to run against that spirit by restricting what can be done with it." Salvatore Sanfilippo (antirez), creator of Redis, published a broader defense of AI reimplementation based on copyright law and GNU project history.
The Core Question: Legal vs Legitimate
The article argues that both Ronacher and Sanfilippo evade the central question: does legal mean legitimate? The author notes that while AI-assisted reimplementation occupies the same legal ground as historical reimplementations (like GNU reimplementing UNIX), there's a crucial directional difference.
When GNU reimplemented UNIX, the vector ran from proprietary to free software, expanding the commons. In the chardet case, software protected by a copyleft license (LGPL) has been reimplemented under a permissive license (MIT) that carries no obligation for derivative works to share their source code. The author argues this removes the fencing that protected the commons, since derivative works built on chardet 7.0 are under no obligation to share their source code.
The article clarifies that the GPL doesn't prohibit keeping source code private or privately modifying GPL software. Its conditions are triggered only by distribution: if you distribute modified code or offer it as a networked service, you must make the source available under the same terms. This creates a condition placed on sharing rather than a restriction on sharing.
📖 Read the full source: HN AI Agents
👀 See Also

AI Learns the 'Dark Art' of RFIC Design — Faster Chips, No Human Intuition Required
Princeton researchers use reinforcement learning and inverse design to create RFICs from scratch. Diffusion models generate novel layouts with record performance, cutting design time drastically.
Claude Code v2.1.210 Fixes Worktree Isolation, Ultracode Opt-In, and Dozens of Bugs
Highlights include subagent worktree isolation fix, ultracode keyword opt-in fix, new elapsed-time counter, and permission rule deprecations.

Meta Releases BOxCrete AI Model for Concrete Mix Design
Meta has released Bayesian Optimization for Concrete (BOxCrete), an open-source AI model for designing sustainable concrete mixes using U.S.-produced materials. The model improves on previous versions with better noise robustness and slump prediction capabilities.
Claude Agent SDK Gets Dedicated Monthly Credits for Programmatic Usage Starting June 15
Starting June 15, paid Claude plans receive a separate monthly credit for programmatic usage (Agent SDK, claude-p, Claude Code GitHub Actions, third-party tools). Pro gets $20, Max 5x $100, etc. Usage pauses if credit runs out and additional usage credits are off.