Bun moving from Zig to Rust feels like a pragmatic engineering decision in the short term, but it might actually end up highlighting Zig’s strengths in the long run (see the discussion on Hacker News: https://news.ycombinator.com/item?id=48016880) .

First point: performance is something you really notice only after you lose it

A big part of why Bun was so fast early on is tightly coupled to Zig’s design—minimal abstraction, explicit control over memory, and very little “hidden cost.” That combination makes it easier to squeeze out performance in real-world hot paths.

Rust can get very close to C/C++ performance, but in practice, the abstractions, borrow checker constraints, and typical idioms often introduce overhead—especially in large, evolving codebases. So I wouldn’t be surprised if we end up in a situation where:

→ the Rust version is more maintainable and ecosystem-friendly → but the old Zig implementation (or parts of it) still outperform it in critical paths

And once you start profiling seriously, that gap tends to become very obvious.

Second point (more interesting): AI changes the calculus of language choice.

As discussed in that Hacker News thread, part of the friction here involves contributing AI-assisted patches upstream. That detail matters more than it seems—it hints at a bigger shift:

👉 code is becoming something you can systematically migrate

If a team can use AI to help move a non-trivial codebase from Zig → Rust, then the reverse is also true:

• Rust → Zig
• Go → Zig
• C++ → Zig

At that point, language choice stops being a one-way door.

Zig is already positioned as a “closer-to-the-metal” language with fewer abstractions. If AI makes large-scale rewrites cheaper, it’s not hard to imagine a pattern emerging:

→ build in Rust (for ecosystem, safety, iteration speed) → rewrite hot paths in Zig (for maximum performance)

Or even more directly:

→ “AI-assisted rewrite to Zig” becomes a standard optimization step

So ironically, this move might not weaken Zig’s position at all.

• Rust wins on ecosystem, tooling, and contributor friendliness
• Zig may increasingly win as the final form for performance-critical code

In a world where codebases are more fluid and AI-assisted migration is viable, Zig doesn’t have to win first—it just has to win last.