Please just try HTMX

Original Article ↗ Hacker News Discussion ↗

Site & TLS / Demo Concerns

  • Early comments focus on the site lacking obvious HTTPS or using a self-signed cert; some say this makes it effectively inaccessible due to browser warnings, others note it now redirects to a valid Let’s Encrypt cert.
  • Several people are confused that the “HTMX POST” demo is actually mocked client-side, not a real server call, which reduces trust in the pitch.

Evangelism, Hype, and Adoption

  • Many are weary of framework evangelism and “just use X” pages; they argue good tech doesn’t automatically win and marketing/influencers matter.
  • Some think the article unfairly attacks npm/React while ignoring that HTMX still runs JS and can coexist with build systems.
  • The HTMX creator appears in the thread, explicitly preferring a “chill vibe,” pointing to more balanced essays and to alternative hypermedia tools like Unpoly.

Where HTMX Works Well (Proponents’ View)

  • Described as “HTML over the wire”: small JS library, attributes on elements trigger HTTP requests, server returns HTML fragments that get swapped into the DOM.
  • Advocates say it excels for CRUD apps, dashboards, intranet tools, admin panels, search/autocomplete, and “forms + tables + lists” where most logic and state live on the server.
  • Benefits cited: no SPA build step, tiny payload vs typical SPA bundles, simpler mental model, reuse of server-side templates, good performance and Lighthouse scores, and easy progressive enhancement.
  • Some report successful production use with Flask, Django, Rails-style stacks, and Go, often combined with Alpine.js or Turbo-like tools.

Critiques: Complexity, State, and Coupling

  • Several who “tried HTMX seriously” say larger apps became brittle: many partial templates, out-of-band swaps, and multiple HTML variants per endpoint made it hard to reason about state and data flow.
  • Complaints include: backend must know detailed frontend structure, implicit coupling via IDs/selectors, difficult error handling, and poor fit for complex client-side state (wizards, drag‑and‑drop, local-first apps, diagram editors, chat, etc.).
  • Some argue React/Vue-style SPAs better encapsulate UI state and behavior, make reuse of JSON APIs easier (esp. for mobile), and avoid mixing rendering logic across client and server.

Alternatives, AI, and Ecosystem

  • Other hypermedia or “HTML over the wire” options discussed: Turbo/Hotwire, Unpoly, Datastar, server-sent partials in various frameworks.
  • A recurring theme: AI/LLM support. Multiple commenters say React/Next.js win in 2024–2025 because LLMs know that ecosystem well, making them more productive than with HTMX/Unpoly, despite HTMX’s conceptual simplicity.
  • Some worry that LLM-driven choices will lock the industry into complex SPA stacks even when simpler hypermedia solutions might suffice.