Open‑Source ‘Civil War’: Bun’s AI‑First Vision vs Zig’s Anti‑AI Stance

Amid the AI programming wave, a deepening rift between open‑source spirit and AI ethics has emerged.

The flashpoint came when Bun forked the Zig language compiler, adding parallel analysis and other optimizations that increased its debug‑build speed by four times.

"We are not planning to submit these improvements upstream because Zig has a strict ban on any contributions created by large language models (LLMs)."

Jarred Sumner, Bun’s founder, then posted a bold prediction: the open‑source community will eventually forbid human contributions, calling human‑written “slop” a nostalgic relic by 2025‑26.

Zig’s community defends the traditional meaning of “contribution,” emphasizing that a pull request represents the contributor’s effort, thought, and cultural alignment, and that AI‑generated code burdens reviewers with low‑quality “slop.”

Sumner counters that AI will reshape production, stating that AI will strictly follow style guides, write exhaustive tests and documentation, study competitor implementations, and iterate tirelessly until all third‑party consistency tests pass.

"My guess is that the open‑source community will split into two camps: Pro‑AI and Anti‑AI."

Ladybird author Andreas Kling expands this vision, listing two consequences for the Pro‑AI side: rapid project outpacing and developers forking or rewriting slower Anti‑AI projects.

Pro‑AI projects will leverage powerful harnesses and automated evaluations (Evals) to surpass human limits.

Anti‑AI advocates argue that software engineering’s core qualities—taste, insight, intuition—cannot be simulated by AI and that unchecked AI code will lead to aesthetic decline and intellectual atrophy.

The debate attracted high‑profile reactions: Marc Andreessen gave a +1 to Sumner’s “future without humans” stance; Eric S. Raymond mocked anti‑AI supporters as “idiots.” A game‑engine developer questioned the evidence of quality improvement, while AI critic Gary Marcus warned that AI looks impressive only to the uninformed.

When AI can generate 90% of code, the question arises who defines the remaining 10% of quality. Sumner answers “Agents,” asserting that automated testing, strict type systems, and tireless agents will guarantee code quality, marginalizing human roles.

Darren Shepherd (Rancher co‑founder) argues that AI‑assisted code will be far higher quality than purely human‑written code.

Survival guidelines for engineers:

Beware of ideological bias; evaluate decisions based on project stage and community culture.

Do not become an AI mouthpiece; build strong guardrails, design rigorous evaluation systems, and act as an architect defining system boundaries.

Accept that the community may split: an elite Anti‑AI enclave preserving human wisdom, and a fast‑moving Pro‑AI frontier driving rapid innovation but also chaos.

In conclusion, tools will become obsolete, but the wisdom and aesthetic judgment of engineering endure.