Building and shipping Mac and iOS apps without opening Xcode

Original Article ↗ Hacker News Discussion ↗

Developers are increasingly automating macOS and iOS build, signing, and release workflows from the command line, often orchestrated by AI coding agents, so they rarely or never touch Xcode’s GUI. Commenters compare bespoke scripts and newer tools (fastlane, Expo, Axiom, strudel, Sweetpad, etc.) to Apple’s traditional toolchain, weighing reduced dependencies and flexibility against the benefits of mature community solutions. Alongside enthusiasm for this “vibe coding” style, many raise concerns about security, reliance on cloud LLMs with sensitive credentials, and Apple’s hardware and platform lock-in.

LLM‑Driven Mac/iOS Development Workflows

  • Many commenters report using AI coding agents (often via CLI) to build, sign, notarize, and ship Mac/iOS apps with minimal or no manual Xcode use.
  • Some say they now mostly act as “hands” for the AI, including for CI pipelines and multi-platform builds.
  • Others highlight that AI handles Apple’s complex tooling surprisingly well, especially when guided by existing scripts or skills.

Xcode vs. Command Line / Existing Automation

  • Several note that CLI‑based builds (xcodebuild, notarytool, etc.) have been standard in CI for years; what’s new is delegating that setup to LLMs.
  • There is disagreement on “never opening Xcode”:
    • Some claim they only needed it once for certs/provisioning or simulators.
    • Others insist some steps (certs, certain provisioning, some debugging) still require the GUI.
  • Xcode’s quality is debated: some find it bloated and crash‑prone; others consider it a solid IDE, especially in newer versions with better AI tooling (MCP, simulator control).

Alternative Tooling and Ecosystems

  • Many suggest mature tools like fastlane, Expo (React Native), Flutter, Tuist, XcodeGen, Sweetpad, Strudel, Axiom, and others that already solve much of this without bespoke scripts.
  • Some worry LLMs encourage one‑off pipelines instead of improving shared tools; others argue bespoke, LLM‑generated scripts reduce dependency overhead.

Security & Privacy Concerns

  • Strong concern about giving LLMs access to source, certs, and SSH keys, especially given past leaks and agent mistakes.
  • Mitigations discussed: separate user accounts, VMs/containers, sandbox harnesses, strict filesystem permissions, secure enclave key storage, and gating SSH actions via TouchID or manual approval.

Platform Lock‑In and Access

  • Several criticize Apple for requiring Macs and paid accounts for realistic iOS development, seeing it as rent extraction and a barrier for low‑budget developers.
  • Others respond that platform owners commonly tie toolchains to their hardware/OS and that low‑end Macs are now relatively affordable.

Debugging, Testing & App Quality

  • Debugging remains a weak spot for fully headless workflows; physical‑device‑only bugs often still need Xcode/LLDB.
  • Some worry that making it trivial to ship apps via AI will increase low‑quality “slop” in the App Store; others counter that tastefully avoiding Xcode doesn’t imply low‑quality apps.