I built the same app 10 times: Evaluating frameworks for mobile performance

Original Article ↗ Hacker News Discussion ↗

Svelte, Vue, and Dev Experience

  • Several commenters strongly endorse Svelte/SvelteKit for simplicity and “easy mode” development; some use Svelte web components to integrate into existing stacks and like the portability.
  • Vue/Nuxt is praised as a balanced, intuitive choice with options vs composition API and good long‑term prospects; some firms deliberately chose it over React for perceived lower “bloat” and better patterns for AI-assisted coding.

React, Next.js, and “Bloat” Debate

  • One camp argues React has become bloated and confusing for newcomers, with multiple “wrong” ways to do things and hook-related footguns (dependency arrays, stale closures).
  • Defenders say the core API hasn’t meaningfully changed besides hooks, React 19 was frictionless, and performance differences in the article are modest; they distinguish React from Next.js and blame the latter more.
  • React Compiler and directives like use no memo are cited by critics as evidence of growing complexity; others see them as minor escape hatches.

Bundle Size, Mobile Performance, and Real-World Networks

  • Many agree the article’s focus on mobile and startup performance is valuable; others think it overstates the practical impact of ~100–150 kB differences, especially on modern networks.
  • A long subthread contests whether 13 kB vs ~500 kB really costs “seconds” in practice; critics call those claims unrealistic, supporters counter with experiences on rural, congested, or throttled connections and refer to known conversion losses from slow loads.
  • Some emphasize JS is far more expensive than images due to parse/execute and main-thread blocking.

Native Apps vs Web and App Store Constraints

  • Some are surprised native apps weren’t benchmarked; others argue the web’s single codebase and instant access outweigh small native speed gains for many business apps.
  • App stores are criticized as “technofeudal”: fees, policy lock‑in, and revocation risk. Others argue for native or hybrid (Capacitor/React Native/Flutter) when offline capability and reliability matter more than deployment simplicity.

Resumable / Next-Gen Frameworks

  • Marko and Solid’s strong metrics are noted; commenters highlight resumability (Marko, Qwik) and islands architectures as more important than raw bundle size alone for time‑to‑interactive.

DX, Ecosystems, and Pragmatic Choices

  • Several developers prioritize familiarity, job market, and ecosystem (React/Next, Django+React, Angular, Quasar) over small performance wins. Some feel “framework jaded” and stick with what they ship fastest in.
  • Others note Next.js DX has worsened compared to Vite-based stacks.

Article Writing and Methodology

  • Mixed reactions to the writing: some find it excellent and engaging; others see repetition, overlong word count, dramatic lines (“technofeudalism”), and claim it’s “AI-slop.”
  • Debate over whether AI assistance invalidates the results; some doubt all 10 implementations were truly expert-reviewed, and question how much a trivial kanban prototype says about large production apps.