[スポンサー] Zed、21世紀のニーズに応えるサンセリフフォントスーパーファミリー
Zedは「読者に本当に必要なものは何か」という問いから開発されたタイプシステム。フランスの眼科病院での視覚障害者テストでは、Zed TextがHelveticaを全患者グループで読書速度で上回った。TextとDisplayの2つの光学バージョン、4つの可変軸、547言語(絶滅危惧言語含む)をサポート。
Zedは「読者に本当に必要なものは何か」という問いから開発されたタイプシステム。フランスの眼科病院での視覚障害者テストでは、Zed TextがHelveticaを全患者グループで読書速度で上回った。TextとDisplayの2つの光学バージョン、4つの可変軸、547言語(絶滅危惧言語含む)をサポート。
Zig's strict anti-LLM contribution policy (banning LLM use for issues, PRs, and comments) is rooted in valuing long-term human contributors over code. Loris Cro explains that reviewing LLM-written PRs does not help grow new trusted project members. As a consequence, Bun—a major Zig-based project—will not upstream performance improvements due to this ban.
Zig creator Andrew Kelley argues that it's a misconception people can't tell who is using LLMs, claiming LLM hallucinations differ fundamentally from human mistakes and that frequent agentic-coded contributors have a noticeable "digital smell." He compares it to smokers entering a room and states his project will not accept LLM-assisted contributions.
The Zig programming language project has adopted a strict anti-AI policy for contributions, requiring all code, comments, and documentation to be human-authored without assistance from AI tools. The project's rationale includes concerns about copyright, code quality, and maintaining a human-centric development culture.
The author details their journey of writing a C compiler in the Zig programming language, covering the implementation of tokenization, parsing, and code generation for a subset of the C language.
The article introduces "contributor poker," a game exploring contributor dynamics in open-source projects, and discusses Zig's decision to ban AI-generated contributions due to concerns about code quality, legal issues, and the value of human participation.