{
  "slug": "project-vend",
  "name": "Anthropic Project Vend (Claudius)",
  "url": "https://www.anthropic.com/research/project-vend-1",
  "category": "experiments",
  "tier": "experiment",
  "one_liner": "Claude ran real office shops; phase one lost money, phase two cleared most loss weeks and booked $2.6k revenue.",
  "what_the_agent_does": "Claudius (Claude) priced inventory, ordered stock via helpers, talked to customers on Slack, and ran mini-fridge shops\u2014later with CRM, CEO agent, and merch agent Clothius.",
  "business_model": "Research experiment (not a commercial product)",
  "metrics": [
    {
      "label": "Profit/Loss",
      "value_usd": null,
      "value_text": "Phase 1: not profitable; sold below cost, excessive discounts, identity crisis (Mar\u2013Apr 2025)",
      "as_of": "2025-06",
      "source_url": "https://www.anthropic.com/research/project-vend-1",
      "source_name": "Anthropic",
      "age_days": 488,
      "older_figure": false,
      "source_sync_stale_since": null,
      "superseded": true,
      "link_check": {
        "status": "ok",
        "code": 200,
        "checked": "2026-10-16"
      },
      "verification": null
    },
    {
      "label": "Profit/Loss",
      "value_usd": null,
      "value_text": "Phase 2: weeks with negative profit margin largely eliminated after CRM, price visibility, procedures, CEO agent, and merch agent (still needed humans)",
      "as_of": "2025-12",
      "source_url": "https://www.anthropic.com/research/project-vend-2",
      "source_name": "Anthropic",
      "age_days": 305,
      "older_figure": false,
      "source_sync_stale_since": null,
      "superseded": false,
      "link_check": {
        "status": "ok",
        "code": 200,
        "checked": "2026-10-16"
      },
      "verification": null
    },
    {
      "label": "Revenue",
      "value_usd": 2649.2,
      "value_text": "Phase 2 Q3 mission progress: $2,649.20 cumulative revenue (17.7% of $15,000 target) across SF/NYC/London shops \u2014 not annualized SaaS ARR",
      "as_of": "2025-12",
      "source_url": "https://www.anthropic.com/research/project-vend-2",
      "source_name": "Anthropic (Seymour Cash status in phase-2 report)",
      "exclude_from_rank": true,
      "age_days": 305,
      "older_figure": false,
      "source_sync_stale_since": null,
      "superseded": false,
      "link_check": {
        "status": "ok",
        "code": 200,
        "checked": "2026-10-16"
      },
      "verification": null
    },
    {
      "label": "Locations",
      "value_usd": null,
      "value_count": 3,
      "value_text": "Expanded to three cities (San Francisco with a second machine, New York, London)",
      "as_of": "2025-12",
      "source_url": "https://www.anthropic.com/research/project-vend-2",
      "source_name": "Anthropic",
      "age_days": 305,
      "older_figure": false,
      "source_sync_stale_since": null,
      "superseded": false,
      "link_check": {
        "status": "ok",
        "code": 200,
        "checked": "2026-10-16"
      },
      "verification": null
    }
  ],
  "founded": 2025,
  "team_size_text": "Andon Labs + Anthropic research collaboration",
  "lessons": [
    "Helpfulness training makes agents easy to talk into unprofitable discounts.",
    "Scaffolding (CRM, procedures, price visibility) mattered more than model alone.",
    "Multi-agent 'CEO' cut discounts ~80% but still authorized many giveaways and wandered into all-night chats.",
    "Concrete revenue ($2.6k vs $15k target) shows agent shops can improve without becoming robust businesses."
  ],
  "tags": [
    "anthropic",
    "andon-labs",
    "vending",
    "autonomy"
  ],
  "last_verified": "2026-10-16",
  "agentic_origin": "agent-native",
  "url_profile": "https://mrrledger.com/b/project-vend/",
  "annualized_revenue_usd": null,
  "annualized_revenue_basis": null,
  "confidence": null,
  "source_sync_stale_since": null,
  "older_figure": false,
  "verified_corroboration": null,
  "headline_verification": null,
  "age_days": null,
  "tier_label": "Experiment",
  "tier_note": "Agent-run or research experiment rather than a conventional company.",
  "independent": true,
  "ownership": null,
  "ownership_watch": null,
  "last_ownership_check": "2026-10-16",
  "revenue_disclosed": false,
  "badge_url": "https://mrrledger.com/badge/project-vend.svg",
  "badge_plain_url": "https://mrrledger.com/badge/project-vend-plain.svg",
  "growth_series": [],
  "related": [
    {
      "slug": "botto",
      "name": "Botto",
      "category": "experiments",
      "tier": "experiment",
      "category_label": "Agent-run experiments",
      "tier_label": "Experiment",
      "one_liner": "Decentralized autonomous AI artist that creates weekly and sells its own work at auction, steered by a voting DAO.",
      "annualized_revenue_usd": 1843484,
      "revenue_disclosed": true,
      "confidence": null,
      "source_sync_stale_since": null,
      "url_profile": "https://mrrledger.com/b/botto/",
      "compare_url": "https://mrrledger.com/compare/?a=project-vend&b=botto",
      "score": 6,
      "why": [
        "same category",
        "same tier",
        "shared: autonomy"
      ]
    },
    {
      "slug": "truth-terminal",
      "name": "Truth Terminal",
      "category": "experiments",
      "tier": "experiment",
      "category_label": "Agent-run experiments",
      "tier_label": "Experiment",
      "one_liner": "Semi-autonomous X bot that accrued crypto via grants and memecoins.",
      "annualized_revenue_usd": null,
      "revenue_disclosed": false,
      "confidence": null,
      "source_sync_stale_since": null,
      "url_profile": "https://mrrledger.com/b/truth-terminal/",
      "compare_url": "https://mrrledger.com/compare/?a=project-vend&b=truth-terminal",
      "score": 6,
      "why": [
        "same category",
        "same tier",
        "shared: autonomy"
      ]
    },
    {
      "slug": "andon-market",
      "name": "Andon Market",
      "category": "experiments",
      "tier": "experiment",
      "category_label": "Agent-run experiments",
      "tier_label": "Experiment",
      "one_liner": "SF brick-and-mortar boutique managed by AI agent Luna (Andon Labs).",
      "annualized_revenue_usd": null,
      "revenue_disclosed": false,
      "confidence": null,
      "source_sync_stale_since": null,
      "url_profile": "https://mrrledger.com/b/andon-market/",
      "compare_url": "https://mrrledger.com/compare/?a=project-vend&b=andon-market",
      "score": 5,
      "why": [
        "same category",
        "same tier"
      ]
    }
  ]
}