The Website Specification

Original Article ↗ Hacker News Discussion ↗

A new site, “The Website Specification,” compiles an opinionated checklist of modern web best practices—from HTML basics and accessibility to HTTP headers, SEO and emerging “agent‑readiness” features like llms.txt. Commenters are split: some find it a useful one‑stop reference or prompt for LLM‑driven audits, while others criticize it as AI‑generated “slop,” overengineered, and too focused on catering to bots rather than humans. The debate also surfaces broader tensions about the complexity of today’s JavaScript‑heavy web, the value of generic checklists, and whether standards for AI agents will prove transient or harmful.

Overall reception of the spec

  • Many see it as a useful, opinionated checklist aggregating scattered standards (HTML, accessibility, headers, structured data) into one place.
  • Others argue it’s overlong, overdesigned, and unlikely to be read or fully implemented; some worry a 100+ item “spec” will intimidate or burden developers.
  • Some commenters used it immediately as an audit guide for their own sites and found it practically helpful.

AI‑generation and “slop” concerns

  • Multiple people inspect the Git history and prose and conclude large parts are LLM‑generated, calling it “AI slop” or “soulless,” even while conceding that the technical content is mostly correct.
  • A few prefer that AI‑assisted content be transparently labeled, but also note current “slop detectors” misclassify clearly human‑written pages.
  • There’s disagreement: some dismiss the site primarily because it’s LLM‑assisted; others say usefulness matters more than authorship.

Agent readiness and llms.txt

  • The “agent readiness” and llms.txt ideas are highly contentious.
  • Supporters:
    • See value in a text/markdown or dedicated agent-friendly representation to cut token costs and bypass JS, auth, and ad cruft.
    • Treat it like modern SEO or CLI ergonomics: not essential, but helpful for current agents.
  • Critics:
    • Argue no major AI providers honor llms.txt, creating a false sense of control and negative ROI.
    • Expect sites to abuse separate agent views (as with search engine cloaking), so serious agents will ignore them.
    • Note that making sites more accessible and semantically structured would inherently make them more agent‑friendly without extra formats.
    • Some see “agent readiness” as a transient fad, comparable to past hype cycles.

Checklists, standards, and misuse

  • Several praise checklists as effective tools, especially for beginners and as memory aids.
  • Others warn generic lists can promote cargo‑culting, over‑engineering, and bureaucratic “Jira stories” enforcing low‑value items.
  • The required/recommended/optional tagging is viewed as an attempt to balance pragmatism with completeness, though some dispute what is labeled “required.”

Web bloat, nostalgia, and implementation gaps

  • Many lament modern JS‑heavy, ad‑ridden sites and nostalgically praise simpler HTML/CSS (and even table layouts, Flash, or HTML4) for speed and readability.
  • Some note the spec’s own site appears CPU‑heavy, lacks caching, and doesn’t fully implement its own “required” items, which undermines its authority.
  • There’s side discussion of .well-known URLs, security.txt, change-password endpoints, login form best practices, and legal/accessibility requirements as concrete areas where such a spec can help.