Zig Bans AI PRs to Bet on Contributors, Not Code
Zig rejects LLM-generated contributions to invest review time in mentoring new contributors into long-term project assets, treating PRs as 'contributor poker' where you bet on the person over perfect code.
Treat Contributors as Long-Term Investments
Successful open-source projects like Zig drown in PRs but prioritize growing new contributors over cherry-picking perfect code. Zig's core team reviews even imperfect first PRs to build confident, prolific contributors—each one an 'investment' that pays off over time. This 'contributor poker' mindset means betting on the person, not the PR contents, much like poker where you play the opponent over the cards. Evidence: Zig's strict policy bans LLMs entirely for issues, PRs, and comments, encouraging native-language posts with human translation.
Contrast with Bun, Zig's most prominent project, acquired by Anthropic in December 2025. Bun's Zig fork added parallel semantic analysis and multiple codegen units to LLVM backend, yielding 4x faster compiles (see diff: oven-sh/zig upgrade-0.15.2…upgrade-0.15.2-fast). Bun won't upstream due to Zig's ban, as LLM assistance dominates their workflow.
LLM Assistance Wastes Maintainer Time
Reviewing LLM-generated PRs yields zero return on investment for maintainers. A perfect AI-written PR doesn't teach the submitter Zig's nuances, so the review effort doesn't create a skilled future contributor. Instead, maintainers could prompt their own LLM to solve the issue faster. Loris Cro (Zig Software Foundation VP of Community) argues this in 'Contributor Poker and Zig's AI Ban': helping imperfect human PRs maximizes ROI by expanding the trusted contributor pool, while AI PRs just demand review without upside.
Trade-off: Strict bans deter quick wins (e.g., Bun's 4x perf) but sustain a human-driven community. Bun's heavy AI use highlights the fork risk—projects split when policies clash with AI workflows.