Data appendix.Counted, not estimated.
One person’s digital exhaust. All of it, itemized.
The twin is built from platform exports one subject actually downloaded — 1,292,700 events across 66 source keys, each tagged with the consent tier an ordinary person would have to grant to hand it over. What was deliberately left out matters as much as what went in.
The corpusRead from ingest 2026-08-12
1,292,700
L0 events
66
source keys
15
platforms
2003 – 2026
coverage span
Events by consent tier
T1 public 62,138 4.8%
T2 behavioral 386,692 29.9%
T3 platform DMs 668,220 51.7%
T4 email & private 175,650 13.6%
Source inventory66 keys · 15 platforms
Every source, every key. Grouped by where it came from.
Each platform export lands as one or more source keys — a DM lane, a like lane, an inference lane. Counts are floors where the platform deletes by design, and every bar shares one linear scale, so hairlines are real but small. The raw table is downloadable below.
| source key | what it is | tier | coverage | events | share of max |
|---|---|---|---|---|---|
| Facebook / Messenger10 keys | 508,909 | 39.4% | |||
| fb_dm | Messenger DMs — HTML export, identity rebuilt per message | T3 | 2003-01-01 – 2026-08-04 | 479,051 | |
| fb_inference | Meta's ad-interest inferences about the subject | T2 | 2013-05-24 – 2026-03-18 | 14,064 | |
| fb_reaction | Likes and typed reactions (type from the icon) | T1 | 1970-01-01 – 2025-04-12 | 5,712 | |
| fb_comment | Comments on others' posts, incl. in groups | T1 | 2009-09-03 – 2021-02-14 | 4,609 | |
| fb_friend | Friend graph: adds, requests sent/received/rejected, removals | T2 | 2009-08-12 – 2026-04-16 | 2,870 | |
| fb_offsite | Off-Facebook activity (third-party site tracking) | T2 | 2024-11-03 – 2026-08-07 | 838 | |
| fb_event | Events hosted, invitations received, RSVP responses | T2 | 2010-10-31 – 2022-06-25 | 664 | |
| fb_post | Own posts, check-ins, and album captions | T1 | 2009-09-03 – 2024-10-02 | 588 | |
| fb_search | On-platform search history | T2 | 2025-09-02 – 2026-08-06 | 278 | |
| fb_group | Group posts and membership activity | T1 | 2010-05-06 – 2024-12-28 | 235 | |
| Instagram8 keys | 194,347 | 15.0% | |||
| ig_dm | Instagram DMs — 1,740 threads, mojibake repaired | T3 | 2013-12-12 – 2026-08-05 | 154,325 | |
| ig_like | Instagram likes | T1 | 2013-03-02 – 2026-08-05 | 16,547 | |
| ig_follow | Followers and following | T1 | 2013-02-27 – 2026-08-04 | 11,787 | |
| ig_story_interaction | Story replies: answered questions, polls, quizzes, story likes | T3 | 2017-12-05 – 2026-08-03 | 3,331 | |
| ig_story_view | Stories viewed (rolling window) | T2 | 2026-07-06 – 2026-08-05 | 3,204 | |
| ig_collection | Saved posts / collections | T1 | undated | 2,630 | |
| ig_comment | Comments on posts and reels | T1 | 2013-02-28 – 2026-07-29 | 2,483 | |
| ig_post | Own posts | T1 | 2014-02-06 – 2024-10-02 | 40 | |
| Gmail mbox (both accounts)1 key | 172,947 | 13.4% | |||
| gmail | All mail incl. spam/trash, both accounts; quotes stripped, is_automated tagged | T4 | 2011-12-05 – 2026-08-05 | 172,947 | |
| Spotify1 key | 134,293 | 10.4% | |||
| spotify | Extended streaming history — plays, skips, conn_country location | T2 | 2011-09-26 – 2026-07-09 | 134,293 | |
| Google Takeout (work + personal)12 keys | 110,561 | 8.6% | |||
| search | Google Search, incl. Discover, News, and Ads activity | T2 | 2012-01-23 – 2071-03-20 | 68,887 | |
| maps | Google Maps activity | T2 | 2012-01-27 – 2026-08-05 | 13,145 | |
| youtube | YouTube watch + search activity | T2 | 2011-12-05 – 2026-08-05 | 9,723 | |
| image_search | Google Image Search | T2 | 2012-01-23 – 2026-08-04 | 4,948 | |
| calendar | Google Calendar events (ATTENDEE = join key) | T2 | 2013-09-26 – 2027-07-24 | 4,435 | |
| gchat | Google Chat messages | T3 | 2015-06-09 – 2026-07-12 | 2,775 | |
| contact | Google Contacts (phone→name = join key) | T2 | undated | 2,631 | |
| shopping | MERGED KEY — Google Play/Shopping + DoorDash orders + Uber Eats | T2 | 2012-01-25 – 2026-08-05 | 1,643 | |
| video_search | Google Video Search | T2 | 2012-02-06 – 2022-01-19 | 775 | |
| flights | Google Flights activity | T2 | 2013-04-06 – 2026-08-05 | 630 | |
| gemini | Gemini Apps conversations | T2 | 2025-03-23 – 2026-05-12 | 563 | |
| hotels | Google Hotels activity | T2 | 2015-09-25 – 2026-08-04 | 406 | |
| Local browsers (device)2 keys | 100,622 | 7.8% | |||
| browser | Local Chrome/Brave/Firefox/Safari history — acquisition=device_local | T2 | 2026-03-24 – 2026-08-07 | 98,754 | |
| browser_search | Search queries parsed out of local browser history | T2 | 2026-03-25 – 2026-08-07 | 1,868 | |
| LinkedIn7 keys | 17,271 | 1.3% | |||
| li_dm | LinkedIn messages — only professional-register conversation corpus | T3 | 2016-12-02 – 2026-07-30 | 7,597 | |
| li_reaction | LinkedIn reactions | T1 | 2016-12-18 – 2026-08-05 | 5,073 | |
| li_connection | Connection graph with company and position | T1 | 2016-01-18 – 2026-07-31 | 2,804 | |
| li_comment | Comments on LinkedIn posts | T1 | 2017-01-24 – 2026-08-04 | 959 | |
| li_invitation | Connection invitations, sent and received | T3 | 2026-01-11 – 2026-08-05 | 665 | |
| li_post | Own posts and shares | T1 | 2017-09-09 – 2026-08-05 | 144 | |
| li_profile | Profile facts: positions, education, skills | T1 | 2018-01-01 – 2026-01-01 | 29 | |
| Snapchat4 keys | 16,917 | 1.3% | |||
| snap_dm | Saved chat only — a floor, not a measure (saved_only flagged) | T3 | 2014-05-05 – 2026-07-23 | 13,479 | |
| snap_memory | Memories metadata (dated filenames) | T4 | 2015-04-05 – 2023-02-26 | 2,120 | |
| snap_friend | Friends list | T2 | 2012-10-16 – 2025-04-02 | 1,316 | |
| snap_snap | Snap history — 2 rows; snaps delete by design | T2 | 2021-08-18 – 2021-08-19 | 2 | |
| X / Twitter4 keys | 15,495 | 1.2% | |||
| x_dm | X direct messages (ids resolved via follower/following) | T3 | 2019-09-20 – 2026-05-01 | 6,997 | |
| x_follow | Follows/followers | T1 | undated | 3,670 | |
| x_like | Likes | T1 | undated | 3,046 | |
| x_post | Tweets | T1 | 2019-09-24 – 2026-08-04 | 1,782 | |
| Media vision captions2 keys | 12,163 | 0.9% | |||
| photo_caption | Vision (LLaVA) captions of personal photos | T2 | 2015-04-04 – 2022-08-06 | 7,748 | |
| screenshot_caption | Vision captions of screenshots/memes — attention signal | T2 | 2015-06-26 – 2021-02-15 | 4,415 | |
| Apple / iCloud9 keys | 3,686 | 0.3% | |||
| apple_contact | iCloud Contacts | T2 | undated | 1,660 | |
| apple_calendar | iCloud Calendars (.ics events) | T2 | 2011-11-06 – 2026-05-13 | 1,268 | |
| apple_note | iCloud Notes — the only no-audience writing in the corpus | T4 | 2016-09-05 – 2026-08-03 | 583 | |
| apple_bookmark | Safari bookmarks | T2 | undated | 117 | |
| apple_care | AppleCare cases | T2 | undated | 18 | |
| apple_device | Device inventory | T2 | undated | 18 | |
| apple_purchase | Apple.com / App Store purchases | T2 | undated | 16 | |
| apple_push | App install & push records | T2 | undated | 5 | |
| apple_wallet | Wallet activity | T2 | undated | 1 | |
| Venmo1 key | 2,054 | 0.2% | |||
| venmo | Transactions: amount/audience/note; identity block deliberately excluded | T2 | 2016-04-19 – 2026-07-28 | 2,054 | |
| Amazon3 keys | 1,892 | 0.1% | |||
| amazon | Order history — dated physical purchases | T2 | 2023-04-28 – 2026-08-09 | 1,805 | |
| amazon_digital | Digital content orders (apps, ebooks, video) | T2 | 2023-04-28 – 2026-07-26 | 77 | |
| amazon_cart | Cart adds that were never bought | T2 | 2024-12-29 – 2026-08-07 | 10 | |
| Lyft1 key | 855 | 0.1% | |||
| lyft_trip | Lyft rides + bike/scooter trips | T2 | 2021-03-11 – 2026-08-07 | 855 | |
| Uber1 key | 688 | 0.1% | |||
| uber_trip | Uber rides — pickup/dropoff coords; cancelled trips kept | T2 | 2019-01-01 – 2026-06-24 | 688 | |
Not yet in the corpus15 exports identified
What the corpus still can't see. Every export verified, ranked by the gap it fills.
Each row is a “download your data” portal that exists, was verified, and has not been requested yet. Ranked by which weak vertical of the model it would fill — not by size. The same consent envelope applies: one-click exports an ordinary person could hand over.
| source | what it adds | how it arrives | status |
|---|---|---|---|
| ChatGPT + Claude | Unmediated reasoning and self-talk — the weakest interior vertical (emotional patterns) | one-click JSON, link expires 24h | priority |
| TikTok | Watch history, search, comments, DMs — the largest platform absent from the corpus | JSON export, ready in 1–4 days | priority |
| Apple Health | Heart rate, sleep, workouts — the weakest behavioral vertical (health) | on-device XML, instant | priority |
| Google Maps Timeline | The location history the Takeout shipped empty — it moved on-device in 2025 | phone-only JSON export | priority |
| PayPal | 7-year transaction ledger with counterparties — Venmo's missing sibling | instant CSV | priority |
| Kalshi | Decision-under-uncertainty ledger for the vices and risk verticals | per-year CSV, web only | priority |
| Netflix | Viewing plus browse/hover interaction history | DSAR, ≤30 days, link lives 7 | priority |
| Apple Media Services | Apple Music plays with durations, searches; possibly Shazam tags | privacy.apple.com request | priority |
| Kindle + Audible + Prime Video | Session-level reading and listening — attention span and abandonment, absent everywhere else | one Amazon data request | priority |
| Resy + OpenTable | Reservation history for the dining vertical | email-only CCPA request | in flight |
| Discord | Community messages and server graph | data package, ≤30 days | slow lane |
| Interest graph, contents undocumented | settings data request | slow lane | |
| Airbnb + Delta | Stays, bookings, loyalty ledgers for the travel vertical | privacy portals, form-based | slow lane |
| Wearables (Oura-class) | 5-minute-interval body telemetry, whichever device is actually worn | self-serve CSV/JSON | slow lane |
| FanDuel + DraftKings | Bet ledgers for the vices vertical | CSV; full ledger needs an extension | slow lane |
Verified dead ends, so no one re-litigates them: Pocket shut down with its export window (Oct 2025). Max withholds watch history even under a formal request. Hulu and Disney+ run undocumented form flows. Resy, OpenTable, Grubhub, Instacart, and Starbucks have no self-serve portal — email-request only. Plex, SoundCloud, Nintendo, and most airlines ship nothing usable. And the standing exclusions hold: personal messaging, data brokers, and identity records stay out no matter how obtainable they are.
Reading notes
What the counts do and don’t say.
- shopping is a merged key: Google Play/Shopping activity, DoorDash orders, and Uber Eats all write to it. Attribution requires the row’s provenance, not the key.
- Coverage dates past 2026 are platform-supplied timestamps kept as-is — future calendar events and one misdated activity row. Correcting them silently would corrupt the audit trail.
- Undated keys (follows, contacts, saved collections, most Apple lanes) are state exports, not event streams — the platform ships no timestamps for them.
- Snapchat chat counts only what somebody pressed save on: a floor, not a measure of use. Local browser history is device-local acquisition, held apart from one-click exports so ablations can price the bigger ask.
- Deliberate exclusions, because the exclusion is the experiment: personal messaging (iMessage), data-broker files, payment identity blocks, raw IP addresses, and media payloads — only vision captions of media enter, never the media itself.