{
  "format": "hyperize/v1",
  "@id": "https://www.hyperize.ai/en/insights/articles/ai-agents-built-our-websites.json",
  "type": "Article",
  "pageType": "Article",
  "title": "AI agents built our websites. One human made the decisions.",
  "alternativeHeadline": "hyperize.ai and minglabs.com shipped without a design or dev team: a design sandbox, a mechanical port, 34 automated checks, and AI engines as the judge.",
  "url": "https://www.hyperize.ai/en/insights/articles/ai-agents-built-our-websites",
  "alternateLanguage": {
    "de": "https://www.hyperize.ai/de/insights/articles/ai-agents-built-our-websites"
  },
  "inLanguage": "en",
  "author": {
    "name": "Marc Seefelder",
    "url": "https://www.hyperize.ai/en/about#marc"
  },
  "publisher": {
    "@id": "https://www.hyperize.ai/#organization",
    "name": "Hyperize",
    "url": "https://www.hyperize.ai",
    "parentOrganization": "MING Labs"
  },
  "datePublished": "2026-06-11",
  "dateModified": "2026-06-11",
  "nextReview": "2026-12-11",
  "evidenceTier": "proprietary",
  "confidence": "B",
  "summary": "The making-of hyperize.ai and minglabs.com: one founder and a set of AI agents shipped two production websites, 17 days from first commit to launch. 12 page types, 34 automated checks per page, done means exit code 0. Eight days after launch, a logged-out ChatGPT cited the new domain as its first source. Including the part that failed: a ported demo form that dropped every lead for four days.",
  "answer": "hyperize.ai and minglabs.com were launched through an AI-agent-driven production pipeline: one human approved every decision, agents executed the design port and the build, and every page passed 34 automated checks before launch. hyperize.ai went from first commit to live in 17 days.",
  "hook": "Two production websites are live: hyperize.ai and minglabs.com. There was no design team and no dev team [S1]. One founder made the decisions. AI agents did the building: they ported the approved design into plain HTML and checked every page against 34 automated rules before it could ship [S6]. Seventeen days from first commit to launch [S1]. Eight days later, a logged-out ChatGPT cited the new domain as its first source [S2].",
  "pipeline": {
    "stations": [
      {
        "name": "Design sandbox",
        "tool": "Lovable",
        "role": "Visual iteration by one founder until the layout holds. Output is treated as a living specification, not as software."
      },
      {
        "name": "Mechanical port",
        "tool": "Claude Code",
        "role": "Reads the design source, reproduces it 1:1 in Astro. Interpretation banned. Design repo read-only for production."
      },
      {
        "name": "Gates",
        "role": "12 page types with machine-readable rules, 34 automated checks per page, content gate on every push. Done = exit code 0."
      },
      {
        "name": "Launch + measurement",
        "role": "Server-rendered HTML, JSON twin per page, llms.txt. Citation probes log wins and losses with the query class declared."
      }
    ],
    "operatingModel": "One human holds taste, approvals, and the veto. Agents execute. Shared memory lives in a vision document, a backlog, and a decision log with 26 ratified decisions."
  },
  "receipts": [
    {
      "claim": "17 days from first commit to launch",
      "evidence": "First production commit 2026-05-15; hyperize.ai live on Netlify 2026-06-01 with 61 EN URLs in the launch sitemap.",
      "sourceId": "S1"
    },
    {
      "claim": "Two sites, one pipeline",
      "evidence": "minglabs.com live 2026-05-28 through the same process.",
      "sourceId": "S1"
    },
    {
      "claim": "Performance without a dev team",
      "evidence": "Lighthouse mobile: Performance 95, LCP 2.0s, CLS 0. Reproducible via pagespeed.web.dev.",
      "sourceId": "S7"
    },
    {
      "claim": "Cited by AI engines 8 days after launch",
      "evidence": "2026-06-09, logged-out ChatGPT (mobile app, no account), plain buyer question: Hyperize index cited as source 1 of 3, above two arXiv web-agent benchmarks. Same day, vocabulary-led queries: ChatGPT cited the Siemens Energy BrandScore above Wikipedia; Perplexity quoted it 6 times inline.",
      "sourceId": "S2",
      "measurementFrame": "Query classes declared: the logged-out run was non-leading (no Hyperize vocabulary fed); the Siemens Energy and Perplexity wins were vocabulary-led. Capability proof, not ROI proof."
    },
    {
      "claim": "The failure, published",
      "evidence": "Homepage lead form validated input and confirmed success but transmitted nowhere, 2026-06-01 to 2026-06-05. A faithfully ported demo form. Every lead in the window lost. Rewired and live-verified.",
      "sourceId": "S8"
    }
  ],
  "sections": [
    {
      "id": "section-01",
      "title": "The bet: the website is the product."
    },
    {
      "id": "section-02",
      "title": "Design in a sandbox, production in a pipeline."
    },
    {
      "id": "section-03",
      "title": "The prettiest page an AI cannot read."
    },
    {
      "id": "section-04",
      "title": "Quality is an exit code, not an opinion."
    },
    {
      "id": "section-05",
      "title": "Live in 17 days, cited in 8 more."
    },
    {
      "id": "section-06",
      "title": "A website is not a project anymore."
    }
  ],
  "twoSentenceStandout": "A process that works once is an anecdote. A process that works twice is a pipeline.",
  "sources": [
    {
      "id": "S1",
      "publisher": "Hyperize Internal — Build record",
      "title": "hyperize.ai build timeline: first production commit 2026-05-15, go-live on Netlify 2026-06-01, 61 EN URLs in the launch sitemap",
      "date": "2026-06-01",
      "url": "https://www.hyperize.ai",
      "supports": "The 17-day timeline; minglabs.com live 2026-05-28 on the same pipeline; design-tool spend for both sites about €1,300.",
      "type": "internal"
    },
    {
      "id": "S2",
      "publisher": "Hyperize Internal — Citation probe record",
      "title": "Logged-out ChatGPT (mobile app, no account), 2026-06-09: plain buyer question, Hyperize cited as source 1 of 3",
      "date": "2026-06-09",
      "url": "https://www.hyperize.ai/en/dax40-index/brands/sap",
      "supports": "Non-leading query, zero personalization. Retained sources: the Hyperize SAP BrandScore plus two arXiv web-agent benchmarks; Hyperize first.",
      "type": "internal"
    },
    {
      "id": "S3",
      "publisher": "Hyperize Internal — Citation probe record",
      "title": "Siemens Energy BrandScore cited by ChatGPT above Wikipedia and 6 times inline by Perplexity, 2026-06-09",
      "date": "2026-06-09",
      "url": "https://www.hyperize.ai/en/dax40-index/brands/siemens-energy",
      "supports": "Vocabulary-led wins, declared as such: the queries carried vocabulary from our published analysis.",
      "type": "internal"
    },
    {
      "id": "S4",
      "publisher": "Hyperize — BrandScore",
      "title": "BMW — configurator content hidden in Shadow DOM web components",
      "date": "2026-05-24",
      "url": "https://www.hyperize.ai/en/dax40-index/brands/bmw",
      "supports": "A coding agent reads the rendered page body as an empty string: JavaScript-dependent content is invisible to agents that fetch instead of render.",
      "type": "internal"
    },
    {
      "id": "S5",
      "publisher": "Vercel",
      "title": "The rise of the AI crawler — measurement of major AI crawlers' JavaScript rendering",
      "date": "2024-12-17",
      "url": "https://vercel.com/blog/the-rise-of-the-ai-crawler",
      "supports": "Verbatim finding: 'none of the major AI crawlers currently render JavaScript', including GPTBot, ClaudeBot and PerplexityBot. Gemini, on Googlebot infrastructure, is the measured exception.",
      "type": "external"
    },
    {
      "id": "S6",
      "publisher": "Hyperize Internal — Build system",
      "title": "Page QA system: 12 page types, 34 automated checks per page, content gate on every push, 26 ratified decisions",
      "date": "2026-06-11",
      "url": "https://www.hyperize.ai/en/methodology",
      "supports": "Done means exit code 0; the decision log replaces re-litigation across build sessions.",
      "type": "internal"
    },
    {
      "id": "S7",
      "publisher": "Hyperize Internal — Performance record",
      "title": "Lighthouse mobile baseline: Performance 95, LCP 2.0s, CLS 0",
      "date": "2026-05-16",
      "url": "https://www.hyperize.ai",
      "supports": "All Core Web Vitals green; reproducible via pagespeed.web.dev.",
      "type": "internal"
    },
    {
      "id": "S8",
      "publisher": "Hyperize Internal — Incident record",
      "title": "Homepage lead form transmitted nowhere from 2026-06-01 to 2026-06-05; rewired and live-verified",
      "date": "2026-06-05",
      "url": "https://www.hyperize.ai",
      "supports": "The 1:1 port replicated a demo form with a fake success confirmation; every lead in the window was lost.",
      "type": "internal"
    }
  ],
  "relatedSurfaces": [
    {
      "relationship": "about",
      "target": "Agent Surface",
      "atId": "https://www.hyperize.ai/en/methodology/agent-surface#defined-term-agent-surface",
      "url": "https://www.hyperize.ai/en/methodology/agent-surface",
      "note": "The methodology this build exists to prove on its own domain."
    },
    {
      "relationship": "mentions",
      "target": "DAX 40 Agent Success Index",
      "atId": "https://www.hyperize.ai/en/dax40-index#dataset",
      "url": "https://www.hyperize.ai/en/dax40-index",
      "note": "The receipts layer: 34 of 40 brands measured at time of writing."
    },
    {
      "relationship": "isPartOf",
      "target": "Insights",
      "atId": "https://www.hyperize.ai/en/insights#webpage",
      "url": "https://www.hyperize.ai/en/insights",
      "note": "Editorial layer of the Hyperize Agent Surface."
    }
  ],
  "scope": {
    "publishes": [
      "The build pipeline at station altitude (design sandbox, mechanical port, gates, launch)",
      "Verified build receipts (timeline, performance, citation probes with declared query classes)",
      "The operating model (one human decides, agents execute, documents as shared memory)",
      "The published failure (the lead form incident)",
      "Named tools (Lovable, Claude Code, Astro, Netlify)"
    ],
    "doesNotPublish": [
      "Gate criteria or check contents (counts only)",
      "Prompts, templates, or skill definitions",
      "The Agent Success Score derivation",
      "Measurement protocols or the frozen task slate"
    ]
  },
  "engagements": [
    {
      "name": "Methodology — Agent Surface",
      "href": "/en/methodology/agent-surface",
      "role": "The layer the build exists to prove."
    },
    {
      "name": "DAX 40 Agent Success Index",
      "href": "/en/dax40-index",
      "role": "How Europe's biggest brands score on the same surface."
    },
    {
      "name": "Founding program",
      "href": "/en/founding-program",
      "role": "Where this pipeline gets deployed on the client's own domain."
    }
  ]
}