Anthropic acquires Bun

Original Article ↗ Hacker News Discussion ↗

Claude Code, $1B ARR, and UX Warts

  • Several commenters use Claude Code daily and say the $1B ARR figure “tracks,” but others highlight serious TUI bugs (flickering/scrolling, keyboard handling, giant JSON state file).
  • Some are surprised a tool touted as 90%-AI-written code has such obvious defects; others say this reflects the current limits of LLM‑authored apps.

Why Anthropic Bought Bun

  • Claude Code’s current binary is already built with Bun; Anthropic is reducing risk around a core dependency that powers a large revenue stream.
  • Bun’s strengths for Anthropic: single cross‑platform binaries, very fast startup, integrated bundler/test runner, native TypeScript, and a batteries‑included standard library (HTTP, S3-like storage, SQL, etc.).
  • Several see this as positioning for agentic workflows: owning a performant JS runtime for “code interpreter”-style skills and sandboxed code execution near data and models.
  • Others think it’s primarily an acquihire plus de‑risking move rather than a deep product pivot.

Risk, Stability, and VC Dynamics

  • Bun was MIT‑licensed, $0 revenue, but had raised ~$26M and claimed ~4 years of runway; many view the deal as an investor exit before having to prove a monetization story.
  • Some argue tying Bun’s “long‑term stability” to a loss‑making AI lab in a possible bubble is the opposite of stability; others counter that Anthropic’s high and fast‑growing revenue makes Bun safer than as a small VC‑backed startup.
  • Multiple people stress that MIT licensing keeps a fork escape hatch if Anthropic deprioritizes or enshitifies Bun.

Bun vs Node/Deno: Technical Debates

  • Pro‑Bun comments emphasize: much faster installs, quick startup, strong Node/npm compatibility, a big built‑in standard library, and easy full‑stack bundling and single‑file executables.
  • Skeptics raise: memory leaks, segfaults, Docker memory issues, immature Zig/JSC stack vs Rust/V8, and “unfocused” scope.
  • Deno partisans cite its permission model and ecosystem‑level security, but many report Bun handled existing Node projects more smoothly.

AI, Coding, and Ecosystem Reactions

  • Thread rehashes claims that AI will soon write “90–100%” of code; some say they’re already near that in web stacks, others report LLM code is still review‑heavy and often poor.
  • Some are delighted a low‑level devtool like Bun found a lucrative AI home; others vow to drop Bun to avoid entanglement with “big AI” and stick with Node or Deno.