AI agents now make native GUI personal software (SwiftUI, macOS) as buildable as an .emacs config, killing the excuse to tolerate TUI or Electron apps.
Key Takeaways
Author built MDV.app, a macOS native Markdown viewer with SQLite FTS, TOC nav, bookmarks, and text search, in ~30 minutes of interactive Claude time.
Electron persists because native UI talent is scarce; Claude closes that gap by being a competent SwiftUI developer, not just a replacement-level one.
The thesis: AI-assisted native apps follow Emacs culture, personal, bespoke, show-and-tell driven, with prompts mattering more than source code.
Target category is long-tail personal tooling: Markdown viewers, iostat wrappers, bpftrace visualizers, things too niche for App Store economics.
Escape velocity is rare; most output will be forgotten like stale elisp, but the ones that spread carry ideas and prompts, not packaged binaries.
Hacker News Comment Review
Strong consensus that the Emacs analogy lands: personal software as configuration, where the prompts are the shareable artifact, not the repo.
Debate on accessibility: technical users get near-instant wins, but non-experts still need enough domain knowledge to direct and maintain agent output.
Recurring practical wishlist names podcast apps, feed readers, Bluesky clients, recipe managers, and chat as obvious reclaim targets for native rewrites.
Notable Comments
@tptacek: argues learning SwiftUI now is like learning Word really well, outcomes are “within millimeters” whether or not you invest in the skill.
@morpheuskafka: asks whether LLMs can finally automate Figma-to-native-platform conversion for CRUD apps, eliminating React Native and Electron at the spec layer.