Claude Code uses Bun written in Rust now

Original Article ↗ Hacker News Discussion ↗

Bun, the JavaScript runtime recently acquired by Anthropic, has been ported from Zig to Rust using AI agents and is already powering the Claude Code terminal client via an unreleased 1.4 build. Commenters debate whether this marks a turning point in software development—showing large, AI-driven rewrites can be shipped to millions—or a risky marketing stunt that undermines code quality, community governance, and the spirit of open source, especially given the heavy reliance on `unsafe` Rust and AI-generated changes. Others focus on practical impacts: Bun’s canary builds appear to work for Anthropic’s use cases, but frequent updates, opaque versioning, and Claude Code’s UI bugs leave some developers questioning its stability and long‑term trustworthiness.

Bun 1.4 and Claude Code

  • Claude Code has been shipping with Bun 1.4 (Rust rewrite) for weeks, before a tagged public 1.4 release.
  • Some see this as “dogfooding” a canary build on a single controlled product; others say it effectively makes Claude users public beta testers.

Governance, Openness, and Pre‑release Use

  • Concern that, post‑acquisition, governance is opaque: unclear who decides Bun’s direction, whether it’s mostly the corporate owner.
  • Debate over whether using a non‑tagged, possibly private build in a closed-source app violates the “spirit” of FOSS, even if the repo and canary builds are public.
  • Some argue expectations are misplaced: an MIT‑licensed project doesn’t imply community governance.

Language Switch and Community Impact

  • Switch from Zig to Rust is seen by some as killing Bun as a community‑driven Zig flagship and signaling corporate control.
  • Others say language choice is secondary to security, maintenance, and hiring; end‑users mostly care that it works.
  • Zig enthusiasts worry this is a blow to Zig’s perceived viability.

AI‑Driven Rewrite Quality

  • Rewrite was largely an LLM‑assisted Zig→Rust transpilation, then iterated.
  • Supporters: tests + safer language + better tooling make it a net win; real-world usage (Claude Code, canary) hasn’t shown major regressions yet.
  • Skeptics: line‑by‑line, non‑idiomatic Rust with massive unsafe usage likely preserves or adds bugs; maintainability is questionable.

Unsafe Rust and Memory Safety

  • Huge number of unsafe blocks (~14k) alarms some; they argue true Rust safety benefits are not yet realized.
  • Others note original Zig code was “all unsafe” in practice; shrinking unsafe regions and annotating them is already progress, with room for gradual cleanup.

Effect on Contributors and Issues

  • Many pre‑rewrite PRs and Zig‑specific issues became obsolete or were auto‑closed, which some see as disrespectful to contributors and community effort.
  • Project management via bots/agents is criticized as sloppy, with both fixed issues left open and unfixed ones closed.

Claude Code UX and Engineering Choices

  • Multiple reports of TUI rendering glitches, high CPU/RAM use, and occasional segfaults; some users have switched to alternative harnesses.
  • Heavy JS/React TUI stack is widely questioned; critics ask why not a lean native TUI if LLMs supposedly make rewrites cheap.
  • Defenders emphasize shared JS logic across web/desktop/CLI and fast iteration; argue architecture is a business decision that “largely works.”

Broader AI Development Implications

  • Many view the successful large-scale port and deployment as evidence that we’ve entered a new phase where LLMs can handle massive rewrites.
  • Others call it primarily a marketing stunt, worry about “AI slop,” supply‑chain risk, and the environmental cost of endless automated refactoring.