Electrobun 2.0 will be decoupled from Bun due to the Rust rewrite

Original Article ↗ Hacker News Discussion ↗

Electrobun, a TypeScript-focused desktop app toolkit built on the Bun JavaScript runtime, plans to decouple from Bun after its core was rapidly rewritten from Zig to Rust using AI tools. Commenters debate whether large, LLM-generated rewrites can be trusted—citing concerns over minimal human review, extensive `unsafe` Rust usage, and the sudden shift being pushed onto users—versus seeing it as an inevitable direction for software development. Many frame the move as a test case for AI-maintained infrastructure, with some predicting forks of the old Zig-based Bun and others questioning long‑term reliance on AI-driven, megacorp-controlled tooling.

Context: Bun’s Rust Rewrite and Electrobun Reaction

  • Bun was rapidly ported from Zig to Rust using LLMs, reportedly ~1M LOC in weeks, then merged to main.
  • Electrobun, a Bun-based desktop app toolkit, plans to decouple from Bun due to concerns about this rewrite.
  • Some expect someone will fork the Zig version; others argue language choice isn’t a “betrayal” and shouldn’t be treated religiously.

Trust, Release Process, and Communication

  • Many see the main issue as process, not AI per se:
    • Massive PR, merged quickly, used as canary within ~1–2 weeks.
    • Perception of minimal human review, especially around thousands of unsafe Rust calls.
    • Promised transparency (blog posts, details) is viewed as lacking; one AI-generated audit post is cited as insufficient.
  • Suggestions: keep Rust as a long-lived v2 branch, run both versions in parallel for months, and clearly mark 1.x as maintenance-only.

AI-Generated Code: Quality, Review, and Maintainability

  • Critics:
    • Tests passing don’t imply correctness, security, or maintainability.
    • Reviewing 1M lines in days is seen as impossible; AI code review is viewed as immature.
    • Risk of a codebase no human really understands, with ongoing dependence on an external AI provider to maintain it.
  • Supporters / moderates:
    • Note Bun has reportedly used LLMs heavily for ~6 months already.
    • Argue that scale, visibility, and bug volume need context; all large software is buggy.
    • Some compare this to machine-made vs hand-made products: speed alone shouldn’t disqualify code.

Ecosystem and Alternatives (npm, Deno, Electrobun)

  • Some point out a perceived double standard: high outrage at Bun vs continued reliance on npm despite repeated security incidents.
  • Others worry about concentrating runtimes among a few big players (Node/Microsoft, Deno, Bun/Anthropic).
  • Electrobun is discussed as a potential lighter alternative to Electron, though few firsthand experiences are shared.

Broader Reactions and Sentiment

  • For some, this is a “bellwether” or “canary” for AI-written large codebases and 2026-era software practice.
  • Others frame the backlash as partly an anti-AI labor/profession protest, reflecting fears of job loss and megacorp control.
  • Several participants emphasize that, regardless of AI, they wouldn’t trust any core runtime or NumPy-like library that’s effectively been rewritten in weeks and not battle-tested over time.