Sites in ChatGPT

Original Article ↗ Hacker News Discussion ↗

OpenAI’s new ChatGPT Sites feature, which lets users “vibe code” and instantly deploy full-stack websites and simple web apps on OpenAI’s infrastructure, is seen as both a powerful convenience and a direct threat to traditional web builders like Wix, Squarespace, and small dev agencies. Commenters highlight how quickly prototypes and niche tools can now be spun up, but worry this accelerates a broader trend: AI labs moving up the stack, cannibalizing their own developer customers, and pushing software and SaaS toward commoditization. Many expect this to shrink demand for routine coding and design work while increasing the importance of moats such as data, infrastructure control, specialized domains, and trust.

Overall Reaction to ChatGPT Sites

  • Many see Sites as a natural, powerful extension: “type a request, get a hosted site/app at a URL,” eliminating deployment headaches for non-technical users.
  • Others find the examples underwhelming or “AI slop” (awkward copy, odd visuals, basic interactivity), though some demos are considered surprisingly polished and non-obvious as AI-generated.
  • Several users report using Sites for weeks/months and say it’s become part of their prototyping workflow.

Business Strategy, Lock-In & Cannibalization

  • Strong view that OpenAI is moving deliberately up the stack: from “intelligence API” to hosting, plugins, auth, databases, and eventually full apps.
  • Some argue this is necessary to escape API commoditization and build a moat/walled garden (Sign in with ChatGPT, hosted plugins, BYO subscription billing).
  • Others worry OpenAI will “Sherlock” AI-dependent startups and SaaS tools, training on customer workloads and competing directly.
  • Counterpoint: cloud/infra vendors (Cloudflare, others) may provide similar agent-friendly hosting; OpenAI may not retain an enduring moat.

Impact on Developers, SaaS, and Jobs

  • Split views:
    • One camp sees this as another coffin nail for low-complexity web work (brochure sites, basic SaaS wrappers, small agencies).
    • Another argues complex apps still need architecture, security, data modeling, operations, and accountability—far beyond vibe-coded sites.
  • Several developers report huge productivity gains from AI (weeks → hours) and expect significant headcount reductions, especially for “just coding” roles.
  • Some foresee fewer but better-paid roles; others anticipate a harsh “musical chairs” market.

Product Capabilities & Comparisons

  • Sites can host full web apps with Cloudflare backend, plugin integration, scheduled updates, and custom domains; can act as both site and plugin.
  • Compared to Claude Artifacts, Sites is seen as more “full stack” (backend, storage) versus primarily static artifacts.
  • Some say GPT/Codex is cheaper or more capable in coding; others strongly prefer Claude’s UX and reliability.

Quality, Safety, and Alternatives

  • Concerns about abuse: scams, miners, honeypots, given easy free hosting.
  • Many remark on “AI aura” in fonts, layout, and especially copy; some think non-technical users can refine this via prompts, bypassing developers.
  • Interest in open-source/self-hosted analogs and MCP-integrated platforms to avoid vendor lock-in and data/idea “rug pulls.”