the dataroom they
didn't mean to publish.

the smoking gun — revision diff

If the data room were a live backend mirror, the numbers would change smoothly with the underlying data. Instead, headline retention numbers jumped 16 points in one revision, demographics had literal hash placeholders, and the editing cadence is one revision every three minutes for seven hours straight. That's not a backend. That's a deck.

edit cadence — does this look like a backend to you?

# 203 revisions over 97.8 hours
# Average gap between revisions: 1735 seconds (~29 minutes)
# But that average is a lie — it hides two human bursts:

Apr 24 22:02  v1     // skeleton created
                     ⋯ 53-hour pause ⋯
Apr 27 03:29  v21
Apr 27 04:32  v41    // 1 hour later, 20 revisions
Apr 27 05:58  v61    // every ~3 min
Apr 27 07:16  v81
Apr 27 07:56  v101   // 7-hour editing burst
Apr 27 08:30  v121
Apr 27 09:51  v141
Apr 27 10:34  v161   // burst ends
                     ⋯ 36-hour pause ⋯
Apr 28 22:25  v181   // final polish session
Apr 28 23:46  v201   // 90 minutes
Apr 28 23:53  v203   // PUBLISHED — board meeting Apr 29?

# Real backends do not produce this cadence.
# Real humans editing a deck in Notion produce this cadence.

the actual exposure

$ curl -sSL https://series-dataroom-api.ws.coursebite.ai/api/dataroom | jq
{
  "page": "main",
  "data": [
    {
      "title": "Retention",
      "metrics": [
        {
          "title": "Retention (All Time)",
          "value": "62%",
          "description": "D30 Retention",
          "expandedData": [
            { "label": "D1 Retention", "value": "87.1%" },
            { "label": "D7 Retention", "value": "78.9%" },
            { "label": "D14 Retention", "value": "71.3%" },
            { "label": "D30 Retention", "value": "62%" },
            { "label": "D60 Retention", "value": "45.4%" }
          ]
        }
      ]
    },
    // ... 11 more sections ...
    {
      "title": "Reactivation",
      "metrics": [{
        "title": "Reactivation / Migrations",
        "insight": "Migration 1: From 1.5M deployed agents (v1
          unique and dupe accounts included, messy scrambled data)
          → 42K users today. Highlights viral capacity in the past
          that we expect to sustain into the future not that infra
          is scalable/on-prem"
      }]
    }
  ]
}

$ curl -sSL https://leaderboard.series.so/api/leaderboard
{
  "error": "querySrv ENOTFOUND _mongodb._tcp.seriesmain.ieqlr6c.mongodb.net"
}
// ↑ leaks Atlas cluster name + shard ID. Production page broken.