Core room · the page that stays current

Master State

This room holds the engagement's working memory — every number, verdict, hypothesis and open gate, each carrying its date and confidence. The hub holds the doors; depth lives here.

MASTER STATE · delta render 2026-09-03 — carries the 09-02 BigQuery read of MN’s cancellation-reason tables: insights 66–68 (the reason data was never missing, only unread; task completion is a real but secondary cancel signal — an earlier “dominant” draft retracted; the picked list and the free text disagree about error codes), the new cancellation-reason card in the overview, one new hypothesis (churned users with no recorded connection — excluded from outputs until MN defines the field), insight 65 and the agent-readable hypothesis sharpened with their first acceptance fixture (the 08-21 call analysis and transcript now on the shelf), and the first-sprint outcome review pinned to Friday 09-11 (~3 h; 08-28 call); before it, the 2026-08-22 (20:54) render — carries the 08-21 call debrief (user-reported; transcript still finalising): insights 64–65 (one sensitive field is not a whole-system access veto; “agent-friendly” is an acceptance contract, not a file extension), two OPEN hypotheses (lawful data-access perimeter; agent-readable Kairos substrate — current War Room coverage unverified) and the debrief on the shelf; before it, the 2026-08-17 (20:07) render — closes F-T6: the unsupported raw-derived dollar-share label is removed while the measured ≈48 % subscription share stands; also carries the F-T5 wording repair (row 43 "Name the query", found by Codex's verification of the integration; opportunities v0.6.2) and the adjudicated Tris red team of the MN DB audit (insight 63: ACCEPT WITH MATERIAL CORRECTIONS; six surface strings corrected on this page and the board — F-T1 executed-not-"not executed", F-T2 session share not revenue and `untested`, F-T4 ii the $162.3k stamp, F-T5 the 93 % conditional, F-T6 the unsupported dollar-share label, F-T7 the 110 % week; the working-tier chip now says what it means; opportunities v0.6.1), the 08-17 MN DB analytical-fitness audit facts row (plan v1.0 executed, Tiers 0–4; the coverage card above derives from the audit's own COVERAGE.json; the 26.5 % monster now carries sampled EVIDENCE — Lee opened round 5, rounds 5–11 hardened the acceptance chain and the draw-cap rule, Talos ACCEPTED v1.2.10 in round 11 and it executed 08-17: 17,077 future-dated rows found whole-table, two 2 % block draws side by side, ≈ $0.13; still unscored, still 26.5 %), the 08-17 warehouse read after the opposing review (insight 62: churn by tier on a month-start cohort, one list price across tiers, churn labels, the Primer seam), and everything before it — the settled Friday 13:00 Europe/Vilnius cadence, the 08-10 work start, the 08-11 warehouse access + integrity finding, Robertas's September board target, the 08-12 upside-economics distillation, the 08-16 quest map for unscouted territory (insight 61), the 08-16 registry read of the two Lithuanian entities (C1: filed FY2025 revenue €3.10M, loss €298k), and Lee's 08-16 definition of the term "MN DB" (= the BigQuery warehouse, nothing wider) · canon: vault → Kairos — MASTER STATE.md · every new Kairos artifact ships as a room under this door.

Overview — the numbers that matter

Every value carries its date and confidence. Insider figures are private — this page never deploys publicly.

Customer estate · S1 · 9 September 2026

Coverage ledger

What we have read, what remains unknown, and which read comes next.

223 tables · 83 definition gaps

22 warehouse datasets, 509 physical objects. The 223 substantive tables exclude 180 GA4 shards and 106 staging copies. GA4 has three logical families tracked separately. With those families included, the non-staging total is 226.

Content-read coverage is unreconciled. The historical 118/223 query flags are not a read rate. For example, airbyte.user_usage was flagged unread although an August 16 read exists. Today’s Gemini comparison and the September 3 cost-column reads are carried forward.

Next programme reads: S2 — customer voice, session feedback and review prompts. S3 — all 20 KPI workbook tabs. These are queued programme slots on invocation, not running jobs.

Definition test: a separately generated recording can cross-check an observed fact; Product, Finance or Support must own the business rule. Two mirrors of one source are correlated evidence.

Read the 28 room dispositions

App_usage

Objects: 38 physical / 38 in 223
Content-read evidence: At least 3 named objects plus one conditional receipt: App_usage_events (sample + column facts; audit receipt not re-verified in S1), App_usage_cancel_reasons (aggregate/regex), users_churn_reasons (flags); users_app_activity (TRACE-001 source per C3 §3; receipt outside S1 set); exact room total unreconciled. users_churn_reasons and claude.vpn_churns share the retained row count/date; shared lineage is a warning to test, not proved by that signature
Historical scanner: 25/38 flagged yes on 09-05; App_usage_events flagged no, a second scanner false negative conditional on the 08-17 audit receipt
Reader/date/source: Talos audit 2026-08-17; Tris 2026-09-02; C1/U1 and audit COVERAGE.json
Next action: scheduled read — S2 on invocation: sessions.feedback/reasons aggregate first, then screened sample; full cancel free-text codebook. Owner: S2 writer. pseudo_id_funnel: not reading, because shared/reset pseudo IDs require a source identity contract before its 66M-row funnel can support transitions. Remaining reads need their own question.
Denominator/falsifier/expiry: Denominator: 38 physical / 38 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

Google_events

Objects: 5 physical / 5 in 223
Content-read evidence: At least 1: events notificationType by month; other object content reads unreconciled
Historical scanner: 2/5 flagged yes on 09-05
Reader/date/source: Tris, 2026-09-05; C2
Next action: not reading, because notification-source completeness and actual grace settings remain unresolved; google_subscription_lifecycle is deferred until its producer mapping and relation to the notification stream are recovered.
Denominator/falsifier/expiry: Denominator: 5 physical / 5 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

Marketing

Objects: 1 physical / 1 in 223
Content-read evidence: Unknown; 1 historical query flag is not a per-object content receipt
Historical scanner: 1/1 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because S1 has no commissioned acquisition question for this object; locate original query and field list before commissioning exposure.
Denominator/falsifier/expiry: Denominator: 1 physical / 1 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

Mysterium_VPN

Objects: 1 physical / 1 in 223
Content-read evidence: Unknown; 0 positive receipts recovered in this bounded S1 source set / 1 object
Historical scanner: 0/1 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because its purpose and present analytical use are unestablished; first recover table purpose and a decision it can inform.
Denominator/falsifier/expiry: Denominator: 1 physical / 1 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

Test

Objects: 10 physical / 10 in 223
Content-read evidence: Unknown; 0 positive receipts recovered in this bounded S1 source set / 10 objects
Historical scanner: 0/10 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because experimental/stale names do not justify full scans; dark_vpn_earnings is a separate supply-cost lead, outside this VPN journey read.
Denominator/falsifier/expiry: Denominator: 10 physical / 10 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

airbyte

Objects: 10 physical / 10 in 223
Content-read evidence: At least 2: devices_time_from_click_to_connect and user_usage; stored-average and counter profiles only
Historical scanner: 3/10 flagged yes on 09-05
Reader/date/source: Tris 2026-09-05; historical user_usage receipt 2026-08-16; 2026-09-07 correction, actor unestablished in C1; C1/C2
Next action: not reading, because timing/usage profiles already exist; recover producing SQL and counter grain before interpreting averages or recurrence.
Denominator/falsifier/expiry: Denominator: 10 physical / 10 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

airbyte_internal

Objects: 106 physical / 0 in 223
Content-read evidence: Unknown; staging excluded from 223; inventory is not a content read
Historical scanner: outside scan denominator
Reader/date/source: Metadata inventory 2026-09-02; content actor/date unestablished
Next action: not reading, because redundant ingestion copies are outside the substantive denominator; reopen only to resolve a specified lineage/replay defect.
Denominator/falsifier/expiry: Denominator: 106 physical / 0 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

analytics_354300879

Objects: 180 physical / 0 in 223; 3 GA4 logical families (non-additive)
Content-read evidence: 3 families inventoried; content-read object total unknown. first_open/login route is reported, but exact GA4 shard attribution is unestablished here
Historical scanner: outside scan denominator
Reader/date/source: Audit family metadata 2026-08-17; C3 first_open result; content actor/date per shard unknown
Next action: not reading, because S1 is not a GA4 event sweep; first bind first_open/login SQL to exact shards, event windows and identity fields. Full warehouse capture remains excluded.
Denominator/falsifier/expiry: Denominator: 180 physical / 0 in 223; 3 GA4 logical families (non-additive); read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

apify

Objects: 10 physical / 10 in 223
Content-read evidence: Unknown; 0 positive receipts recovered in this bounded S1 source set / 10 objects
Historical scanner: 0/10 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because competitor price scrape is a market comparison, not a customer journey fact; revisit for an explicit price-comparison decision.
Denominator/falsifier/expiry: Denominator: 10 physical / 10 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

claude

Objects: 13 physical / 13 in 223
Content-read evidence: At least 7: users_cycles; vpn_primer_payments; vpn_stripe_charges; vpn_users/purchases/churns/refunds (four S5 comparisons). Selected fields only
Historical scanner: 10/13 flagged yes on 09-05
Reader/date/source: Tris 2026-09-03 cost bridge, 09-05 Primer, 09-09 S5; 09-07 reachability actor unestablished; C2/U1/U2/U3
Next action: not reading, because documented reads already answer the inventory question; excluded vpn_purchases remains outside default analysis authority. Fee/cost lineage and Primer decline-code review are distinct next questions.
Denominator/falsifier/expiry: Denominator: 13 physical / 13 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

gemini

Objects: 5 physical / 5 in 223
Content-read evidence: 5/5 objects aggregate-read in S5: four monthly pair comparisons and traffic date/count inspection; no full-row semantic read claim
Historical scanner: 0/5 flagged yes on 09-05
Reader/date/source: Tris, 2026-09-09, U2 and retained s5-two-truths SQL/JSON
Next action: not reading, because S5 already performed the deferred comparison; inspect transformations/direct Looker configuration next. Similar values or names do not prove independent lineage or downstream wiring.
Denominator/falsifier/expiry: Denominator: 5 physical / 5 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

goproxies_web

Objects: 2 physical / 2 in 223
Content-read evidence: Unknown; 2 historical query flags / 2 objects
Historical scanner: 2/2 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because GoProxies-specific customer analysis is outside this VPN journey programme; preserve its coverage row.
Denominator/falsifier/expiry: Denominator: 2 physical / 2 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

imports

Objects: 1 physical / 1 in 223
Content-read evidence: Unknown; 0 positive receipts recovered in this bounded S1 source set / 1 object; country_tier is inventoried
Historical scanner: 0/1 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because table-name discovery does not commission pricing tiers; use only after version/effective-date and classification rule are recovered.
Denominator/falsifier/expiry: Denominator: 1 physical / 1 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

intercom

Objects: 13 physical / 13 in 223
Content-read evidence: At least 5/13: contacts, conversations, conversation parts, activity_logs, tickets (examined but zero rows in retained inventory); bounded keys/content, not all message bodies
Historical scanner: 2/13 flagged yes on 09-05
Reader/date/source: Six-stream execution 2026-09-03/04 reviewed in C3; per-query actor needs receipt reconciliation; reachability 2026-09-07, actor unestablished
Next action: scheduled read — S2 on invocation: audit existing classified payer conversations. Owner: S2 writer; keep raw bodies in private store and release aggregates only.
Denominator/falsifier/expiry: Denominator: 13 physical / 13 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

intercom_myst_nodes

Objects: 13 physical / 13 in 223
Content-read evidence: At least 3/13: contacts, conversations, conversation parts in six-stream execution; whole-body read count unknown
Historical scanner: 0/13 flagged yes on 09-05
Reader/date/source: Execution 2026-09-03/04, C3; per-query actor needs receipt reconciliation
Next action: not reading, because node-runner support is a distinct supply-side question; raw discovery stays registered, not treated as VPN customer evidence.
Denominator/falsifier/expiry: Denominator: 13 physical / 13 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

intercom_superproxy

Objects: 13 physical / 13 in 223
Content-read evidence: At least 3/13: contacts, conversations, conversation parts in six-stream execution; whole-body read count unknown
Historical scanner: 3/13 flagged yes on 09-05
Reader/date/source: Execution 2026-09-03/04, C3; per-query actor needs receipt reconciliation
Next action: not reading, because proxy support is outside the VPN customer denominator; reuse only for an explicitly scoped proxy question.
Denominator/falsifier/expiry: Denominator: 13 physical / 13 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

looker

Objects: 14 physical / 14 in 223
Content-read evidence: At least 1: users_revenue, 08-17 reconciliation cited by U2; receipt not re-executed. Review-prompt table separately tracked below; exact room total unknown
Historical scanner: 12/14 flagged yes on 09-05
Reader/date/source: C1 register 2026-09-05; U2 cites users_revenue read 2026-08-17, actor unreconciled
Next action: scheduled read — S2 on invocation: app_users_review_prompt positive/negative/feedback-opened counts by period; retain unknown states. Owner: S2 writer. app_users_protocol_events: not reading, because protocol event-to-attempt mapping and success meaning need a producer contract first.
Denominator/falsifier/expiry: Denominator: 14 physical / 14 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

postgre

Objects: 67 physical / 67 in 223
Content-read evidence: At least 1: mv_user_revenue_actual; mv_vpn_cancel_reasons has a historical query flag only pending receipt. Mirrors of claude reads are not independent recordings
Historical scanner: 52/67 flagged yes on 09-05
Reader/date/source: U2 METRIC-CONTRACT and capture-manifest OID bridge; exact content actor/date and room total unreconciled
Next action: not reading, because economics requires per-cycle build lineage and payment-component reconciliation first. omnisend_contacts and omnisend_consent_contacts are deferred until a lifecycle exposure/consent question specifies the intended join and policy; a contact inventory alone cannot establish exposure.
Denominator/falsifier/expiry: Denominator: 67 physical / 67 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

subscriptions

Objects: 2 physical / 2 in 223
Content-read evidence: 2/2: subscription and subscription_events; keys, event types and timestamp availability read, meanings open
Historical scanner: 2/2 flagged yes on 09-05
Reader/date/source: Six-stream execution 2026-09-03/04; reachability 2026-09-07, actor unestablished; C3/U3
Next action: not reading, because joint reachability and chronology profiles exist; next read is event producer/transform contract, including null emitted_at and mutable subscription state.
Denominator/falsifier/expiry: Denominator: 2 physical / 2 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

superproxies

Objects: 2 physical / 2 in 223
Content-read evidence: Unknown; 2 historical query flags / 2 objects
Historical scanner: 2/2 flagged yes on 09-05
Reader/date/source: Actor/date unestablished; C1 register
Next action: not reading, because this is a proxy namespace outside the VPN customer programme; historical flags remain visible.
Denominator/falsifier/expiry: Denominator: 2 physical / 2 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

users

Objects: 2 physical / 2 in 223
Content-read evidence: At least 1/2: creations root frame; root creation is not the full 2026 paying estate
Historical scanner: 2/2 flagged yes on 09-05
Reader/date/source: Six-stream 2026-09-03/04; 2026-09-07 reachability, actor unestablished; C3/U3
Next action: not reading, because the existing frozen-frame read is reusable; a paid-estate question must explicitly include pre-2026-created payers and keep its separate denominator.
Denominator/falsifier/expiry: Denominator: 2 physical / 2 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

vpn_web_tracking

Objects: 1 physical / 1 in 223
Content-read evidence: Unknown; events_recent was excluded from the six-stream run pending lineage, not proved unread across Kairos
Historical scanner: 0/1 flagged yes on 09-05
Reader/date/source: Execution 2026-09-03/04, C3
Next action: not reading, because its source lineage is unresolved under that run contract; recover producer evidence and define a new bounded read before admission.
Denominator/falsifier/expiry: Denominator: 1 physical / 1 in 223; read scope: preceding cell. Falsifier: a dated SQL/result or content-read receipt for additional objects/fields. Expiry: next matching read/lineage receipt or 2026-09-16, whichever comes first.

Drive capture

Objects: 3 selected files; KPI workbook 20 tabs; whole Drive denominator unknown
Content-read evidence: 2 other files content-read; KPI workbook 2/20 tabs reported read (26 VPN MRKT, SP MRKT); 18 tab reads unconfirmed
Historical scanner: not applicable
Reader/date/source: Talos, 2026-09-07 21:43–22:10; capture MANIFEST addendum 09-08
Next action: scheduled read — S3 on invocation: all 20 tabs, formulas/blank/#REF states and target/actual denominators, with blind extraction. Owner: S3 writer.
Denominator/falsifier/expiry: Denominator: 3 selected files; KPI workbook 20 tabs; whole Drive denominator unknown; read scope: preceding cell. Falsifier: a dated additional content-read receipt, including its actual object/field IDs. Expiry: next matching receipt or 2026-09-16, whichever comes first.

ClickUp capture

Objects: 7,068 distinct tasks / 226 lists / 13 spaces; 1,979 comment response files
Content-read evidence: 7,068 task records parsed for capture; semantic task coverage unknown. Six-stream searched 3,605 comments / 2,305 files in its own captured scope; unique overlap with this capture unreconciled
Historical scanner: not applicable
Reader/date/source: Capture complete 2026-09-03 12:48Z; source capture script, agent identity not recorded in receipt; C3 execution 09-03/04
Next action: not reading, because parsing and literal search do not establish semantic understanding; next work should bind a named journey/incident to task and comment IDs before reading relational support links.
Denominator/falsifier/expiry: Denominator: 7,068 distinct tasks / 226 lists / 13 spaces; 1,979 comment response files; read scope: preceding cell. Falsifier: a dated additional content-read receipt, including its actual object/field IDs. Expiry: next matching receipt or 2026-09-16, whichever comes first.

Grafana definitions capture

Objects: 152 dashboard definitions + 17 folder records + 18 datasource definitions in 09-02 capture
Content-read evidence: 152/152 dashboards structurally parsed in prior commissioning; 2,280 panels / 3,133 queries / 286 variables inventoried in 08-27 snapshot. Semantic query-definition coverage unknown; snapshot folder counts differ
Historical scanner: not applicable
Reader/date/source: Talos, 2026-08-27 commissioning; 2026-09-02 capture script (agent not in receipt); U5
Next action: not reading, because a definition inventory is not telemetry or a semantic contract; next read is one source-owned metric expression, unit, retention and producer path for a named outcome.
Denominator/falsifier/expiry: Denominator: 152 dashboard definitions + 17 folder records + 18 datasource definitions in 09-02 capture; read scope: preceding cell. Falsifier: a dated additional content-read receipt, including its actual object/field IDs. Expiry: next matching receipt or 2026-09-16, whichever comes first.

Slack frozen export

Objects: 237 channels / 1,085,280 messages; snapshot ends 2026-08-18
Content-read evidence: 1,085,280 indexed/identifier-searched rows; semantic read total unknown. Contract 2A examined 322 records; the 114-hit fixture is a different bounded retrieval and must not be added
Historical scanner: not applicable
Reader/date/source: Talos/Tris prior analyses 2026-08-24–27; U6/U4 and Operating Map
Next action: not reading, because whole-archive indexing has already run; inspect selected historical lineage/decision threads when an exact field or task provides a retrieval key. Attachments/private channels/later activity have separate unknown denominators.
Denominator/falsifier/expiry: Denominator: 237 channels / 1,085,280 messages; snapshot ends 2026-08-18; read scope: preceding cell. Falsifier: a dated additional content-read receipt, including its actual object/field IDs. Expiry: next matching receipt or 2026-09-16, whichever comes first.

Intercom raw — cross-dataset view

Objects: 39 warehouse objects across 3 workspaces; overlaps intercom* rows above; not additive
Content-read evidence: At least 11 named raw objects in six-stream (3×contacts/conversations/parts plus activity_logs/tickets); 8,606 payer conversations model-classified in programme, blind audit of 339 still open
Historical scanner: not applicable
Reader/date/source: Six-stream 2026-09-03/04; paying-map classification reported in U1 2026-09-08; exact classification actor/date in that run receipt
Next action: scheduled read — S2 on invocation: blind audit, 339 records in batches of 40, against one codebook. Owner: S2 writer; existing classifier output remains provisional.
Denominator/falsifier/expiry: Denominator: 39 warehouse objects across 3 workspaces; overlaps intercom* rows above; not additive; read scope: preceding cell. Falsifier: a dated additional content-read receipt, including its actual object/field IDs. Expiry: next matching receipt or 2026-09-16, whichever comes first.

Review prompt — looker.app_users_review_prompt

Objects: 1 warehouse table / about 241k rows in C1; overlaps looker; not additive
Content-read evidence: Content-read total unknown; inventory/name recognition does not settle positive/negative/feedback-opened counts
Historical scanner: not applicable
Reader/date/source: C1 2026-09-03; S2 pending in U1 2026-09-08; content reader/date unestablished
Next action: scheduled read — S2 on invocation: count prompt state/outcome by period and platform, inspect source event mapping; store-star ratings and prompt-positive are distinct outcomes.
Denominator/falsifier/expiry: Denominator: 1 warehouse table / about 241k rows in C1; overlaps looker; not additive; read scope: preceding cell. Falsifier: a dated additional content-read receipt, including its actual object/field IDs. Expiry: next matching receipt or 2026-09-16, whichever comes first.

Inspect the 83 definition gaps

F001 · source

Stage: Acquisition
Field/fact: source
Meaning gap: Mutable contact attribution versus source-at-time-of-acquisition
Independent recording: Candidate only; independent lineage/record availability unconfirmed: web/app events ↔ affiliate reporting
Resolution/owner: Empirical → business rule — Data + Marketing
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F002 · campaign_id

Stage: Acquisition
Field/fact: campaign_id
Meaning gap: Campaign identity and cross-system namespace
Independent recording: Candidate only; independent lineage/record availability unconfirmed: campaign catalogue ↔ exposure events
Resolution/owner: Empirical → business rule — Marketing + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F003 · creative / offer identifier

Stage: Acquisition
Field/fact: creative / offer identifier
Meaning gap: Version at exposure; a campaign name is insufficient
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical → business rule — Marketing
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F004 · click_id / install_id

Stage: Acquisition
Field/fact: click_id / install_id
Meaning gap: Click-to-install join, resets and collisions
Independent recording: Candidate only; independent lineage/record availability unconfirmed: affiliate click/OID ↔ GA4 install/login
Resolution/owner: Empirical — Data + acquisition source owner
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F005 · exposure_time

Stage: Acquisition
Field/fact: exposure_time
Meaning gap: Event time versus ingest time; actual visibility versus eligibility
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical → business rule — Marketing + experiment owner
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F006 · consent / attribution window

Stage: Acquisition
Field/fact: consent / attribution window
Meaning gap: Permitted window, first/last-touch choice and consent meaning
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Business rule required — Marketing + privacy owner
Next action: Locate adopted policy and consent record; request a rule only for unresolved policy
Source: C4 §5

F007 · visitor_id / user_pseudo_id

Stage: Identity
Field/fact: visitor_id / user_pseudo_id
Meaning gap: Install-scoped identifier, resets and Windows shared identifier
Independent recording: Candidate only; independent lineage/record availability unconfirmed: account login events ↔ source install records
Resolution/owner: Empirical — Client engineering + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C1 timing correction; C3 limits; C4 §4

F008 · user_id / account uuid

Stage: Identity
Field/fact: user_id / account uuid
Meaning gap: Account versus person; namespace, merge/split and effective ownership
Independent recording: Candidate only; independent lineage/record availability unconfirmed: account roots ↔ subscription/support source identifiers
Resolution/owner: Empirical → business rule — Data + identity source owner
Next action: Recover source contract and test keys/windows before promotion
Source: C3; C4 §4; U3

F009 · created_date / account creation

Stage: Identity
Field/fact: created_date / account creation
Meaning gap: Creation event semantics, timezone and left-censored root eligibility
Independent recording: Candidate only; independent lineage/record availability unconfirmed: creation events ↔ account source record
Resolution/owner: Empirical → business rule — Data
Next action: Recover source contract and test keys/windows before promotion
Source: C3 §7; U3

F010 · login method

Stage: Identity
Field/fact: login method
Meaning gap: Account linkage and anonymous-to-known transitions
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical → business rule — Identity source owner
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F011 · device_id

Stage: Identity
Field/fact: device_id
Meaning gap: Device versus install/account; shared/reused values
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client event ↔ device registry
Resolution/owner: Empirical → business rule — Client engineering + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F012 · platform / app_version

Stage: Identity
Field/fact: platform / app_version
Meaning gap: Capture origin, missing values and version validity
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client headers ↔ telemetry dimensions
Resolution/owner: Empirical — Client engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F013 · identity edge / merge history

Stage: Identity
Field/fact: identity edge / merge history
Meaning gap: Cardinality, effective dates, deleted accounts and conflict precedence
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical → business rule — Identity source owner + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F014 · order_id / purchase_id

Stage: Payment
Field/fact: order_id / purchase_id
Meaning gap: Retry deduplication, immutable transaction identity and rail coverage
Independent recording: Candidate only; independent lineage/record availability unconfirmed: gateway events ↔ warehouse transactions
Resolution/owner: Empirical → business rule — Payments + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F015 · payer identifier

Stage: Payment
Field/fact: payer identifier
Meaning gap: One payer/many accounts; account identity is not payer identity
Independent recording: Candidate only; independent lineage/record availability unconfirmed: processor customer ↔ account/subscription mapping
Resolution/owner: Empirical → business rule — Payments + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F016 · subscription_id / user_uuid

Stage: Payment
Field/fact: subscription_id / user_uuid
Meaning gap: Membership versus purchase/entitlement; namespace and orphan events
Independent recording: Candidate only; independent lineage/record availability unconfirmed: subscription source ↔ payment records
Resolution/owner: Empirical → business rule — Subscriptions + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C3; C4 §4; U3

F017 · authorization state

Stage: Payment
Field/fact: authorization state
Meaning gap: Authorized is not captured/settled; transition and reversal rules
Independent recording: Candidate only; independent lineage/record availability unconfirmed: processor ledger ↔ emitted authorization events
Resolution/owner: Empirical → business rule — Payments + Finance
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F018 · capture / SETTLED / succeeded state

Stage: Payment
Field/fact: capture / SETTLED / succeeded state
Meaning gap: State mapping, currency and final cash treatment across rails
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source processor settlement ledger ↔ Primer/Stripe records
Resolution/owner: Empirical → business rule — Payments + Finance
Next action: Recover source contract and test keys/windows before promotion
Source: C2 Primer; U3 Stripe caveat; C4 §3

F019 · currency

Stage: Payment
Field/fact: currency
Meaning gap: Minor-unit exponent, transaction/settlement currency and FX rule
Independent recording: Candidate only; independent lineage/record availability unconfirmed: processor source currency ↔ warehouse field
Resolution/owner: Empirical → business rule — Finance + Data
Next action: Recover source contract and test keys/windows before promotion
Source: U3 Stripe minor-unit caveat; C4 §5

F020 · gross amount / amount_usd

Stage: Payment
Field/fact: gross amount / amount_usd
Meaning gap: Attempt value versus booked amount; FX date, null and duplicate logic
Independent recording: Candidate only; independent lineage/record availability unconfirmed: processor settlement/export ↔ warehouse rows; curated mirrors are correlated
Resolution/owner: Empirical → business rule — Finance + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F021 · refund amount / refund_date

Stage: Payment
Field/fact: refund amount / refund_date
Meaning gap: Refund request versus paid refund; partials, attribution and April layer discrepancy
Independent recording: Candidate only; independent lineage/record availability unconfirmed: refund source event/settlement ↔ modelled refund rows
Resolution/owner: Empirical → business rule — Finance + Payments
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5; U2 (comparison only)

F022 · dispute / chargeback state

Stage: Payment
Field/fact: dispute / chargeback state
Meaning gap: Opened, won/lost and debited are different states; netting scope
Independent recording: Candidate only; independent lineage/record availability unconfirmed: dispute source lifecycle ↔ gateway/warehouse events
Resolution/owner: Empirical → business rule — Finance + Payments
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F023 · gateway/store fee / fee_per_tx

Stage: Payment
Field/fact: gateway/store fee / fee_per_tx
Meaning gap: Actual charged fee versus assumed rate; tax/FX/reversal handling
Independent recording: Candidate only; independent lineage/record availability unconfirmed: processor/store statements ↔ cost model
Resolution/owner: Empirical → business rule — Finance
Next action: Recover source contract and test keys/windows before promotion
Source: C1; U1 cost bridge correction; C4 §5

F024 · payment_error_decline_code

Stage: Payment
Field/fact: payment_error_decline_code
Meaning gap: Issuer/processor code meaning and recoverability; first read skipped codes
Independent recording: Candidate only; independent lineage/record availability unconfirmed: processor response ↔ Primer code field
Resolution/owner: Empirical, then recovery policy — Payments
Next action: Read code/message aggregates; recover provider mapping and policy before recovery claims
Source: C2 Primer

F025 · payment_error_processor_message

Stage: Payment
Field/fact: payment_error_processor_message
Meaning gap: Free-text processor explanation; taxonomy and truncation
Independent recording: Candidate only; independent lineage/record availability unconfirmed: original processor response ↔ Primer message
Resolution/owner: Empirical — Payments
Next action: Inspect screened messages and versioned processor mapping
Source: C2 Primer

F026 · plan / obligation

Stage: Entitlement
Field/fact: plan / obligation
Meaning gap: Plan identity, purchased duration, obligations and offer versions
Independent recording: Candidate only; independent lineage/record availability unconfirmed: product catalogue ↔ purchase/subscription payload
Resolution/owner: Empirical → business rule — Product + Subscriptions
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F027 · entitlement start / end

Stage: Entitlement
Field/fact: entitlement start / end
Meaning gap: Access window versus billing cycle; timezone, overlaps and corrections
Independent recording: Candidate only; independent lineage/record availability unconfirmed: access-control entitlement ↔ billing lifecycle
Resolution/owner: Empirical → business rule — Product + Subscriptions
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F028 · grace / retry / hold

Stage: Entitlement
Field/fact: grace / retry / hold
Meaning gap: Notification absence is not proof a console policy was disabled
Independent recording: Candidate only; independent lineage/record availability unconfirmed: Google RTDN ↔ Play Console configuration/audit history (not captured here)
Resolution/owner: Empirical → business rule — Subscriptions + Play Console owner
Next action: Check source completeness and archived setting history; only human owner settles undocumented policy intent
Source: C2 Google events; C4 §5

F029 · pause / renewal effective date

Stage: Entitlement
Field/fact: pause / renewal effective date
Meaning gap: Requested versus effective state; late arrival and recovery transitions
Independent recording: Candidate only; independent lineage/record availability unconfirmed: store/backend lifecycle ↔ subscription events
Resolution/owner: Empirical → business rule — Subscriptions
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F030 · cancellation effective date

Stage: Entitlement
Field/fact: cancellation effective date
Meaning gap: Customer request versus access ending; voluntary/involuntary mapping
Independent recording: Candidate only; independent lineage/record availability unconfirmed: cancellation source event ↔ entitlement and payment histories
Resolution/owner: Empirical → business rule — Subscriptions + Product
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F031 · subscription_events.event_type

Stage: Chronology
Field/fact: subscription_events.event_type
Meaning gap: Producer meaning of create/update; no conversion implication
Independent recording: Candidate only; independent lineage/record availability unconfirmed: subscription state transitions ↔ source producer code
Resolution/owner: Empirical → business rule — Subscriptions + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C3 §4; U3

F032 · subscription_events.emitted_at

Stage: Chronology
Field/fact: subscription_events.emitted_at
Meaning gap: Mostly null; U3 measured emitted_at = period_start on all 87,569 populated 2026 update rows, so it is not event time and cannot measure delivery lag; timezone and missingness mechanism remain unresolved
Independent recording: Candidate only; independent lineage/record availability unconfirmed: producer event log ↔ receipt event (same event lineage, not independent by default)
Resolution/owner: Empirical — Subscription engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C3 §4; U3 chronology

F033 · subscription_events.received_at

Stage: Chronology
Field/fact: subscription_events.received_at
Meaning gap: Reception semantics, duplicates, ties and backfill; not event order
Independent recording: Candidate only; independent lineage/record availability unconfirmed: ingestion/job logs ↔ source clock
Resolution/owner: Empirical — Data engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C3 §4; U3 chronology

F034 · _airbyte_extracted_at

Stage: Chronology
Field/fact: _airbyte_extracted_at
Meaning gap: Ingestion timestamp versus business observation window
Independent recording: Candidate only; independent lineage/record availability unconfirmed: ingestion run logs ↔ table extraction timestamp
Resolution/owner: Empirical — Data engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C3 bounded-day step 1

F035 · attempt_id / session_id

Stage: Product attempt
Field/fact: attempt_id / session_id
Meaning gap: Eligible attempt grain, retries and session-to-attempt relationship
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client attempt events ↔ server/session records
Resolution/owner: Empirical → business rule — Product + client/backend engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F036 · destination / use-case code

Stage: Product attempt
Field/fact: destination / use-case code
Meaning gap: Customer job representation and allowed collection scope
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Business rule + customer recording required — Product/Research + privacy owner
Next action: Inspect already supplied job evidence; define privacy-safe coding only under source-owner authority
Source: C4 §5

F037 · start / end / event_timestamp

Stage: Product attempt
Field/fact: start / end / event_timestamp
Meaning gap: Client/server clocks, duration unit, poisoned dates and event pairing
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client events ↔ server clock/log
Resolution/owner: Empirical — Client/backend engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C1 timing correction; C3; C4 §5

F038 · error_code

Stage: Product attempt
Field/fact: error_code
Meaning gap: Raw emitted fault versus user survey label; version-specific mapping
Independent recording: Separately authored recording exists; linkage unchecked: separate free-text customer reports exist alongside survey choices; exact attempt linkage unconfirmed
Resolution/owner: Empirical → Product interpretation — Client engineering + Product
Next action: Map actual source codes/version to telemetry, survey choice and independent text; keep unmatched reports
Source: U1 S2; C3 sessions schema; C4 §5

F039 · error_type

Stage: Product attempt
Field/fact: error_type
Meaning gap: Classification producer, nulls and whether multiple errors collapse
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source client/server event ↔ sessions field
Resolution/owner: Empirical — Client engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F040 · status_code

Stage: Product attempt
Field/fact: status_code
Meaning gap: Protocol/client state versus technical/business success
Independent recording: Candidate only; independent lineage/record availability unconfirmed: server/session state ↔ client event
Resolution/owner: Empirical → business rule — Client/backend engineering + Product
Next action: Recover source contract and test keys/windows before promotion
Source: C3 sessions; C4 §3/5

F041 · server / node / path reference

Stage: Product attempt
Field/fact: server / node / path reference
Meaning gap: Network entity and timing; aggregate Grafana label is not account identity
Independent recording: Candidate only; independent lineage/record availability unconfirmed: server/session logs ↔ Grafana series; per-account edge unconfirmed
Resolution/owner: Empirical — Network/backend engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F042 · opened_app_total

Stage: Product attempt
Field/fact: opened_app_total
Meaning gap: False could mean no joined telemetry rather than no open
Independent recording: Candidate only; independent lineage/record availability unconfirmed: raw open events ↔ churn flag, lineage may be shared
Resolution/owner: Empirical → business rule — Data + client engineering
Next action: Recover SQL and compare raw opens by identity/platform/window; retain unknown matches
Source: C4 §10; U1 S4

F043 · was_connected_total

Stage: Product attempt
Field/fact: was_connected_total
Meaning gap: False flag overlaps users_cycles; grain/window/identity uncommissioned
Independent recording: Candidate only; independent lineage/record availability unconfirmed: app_session_connections ↔ churn flag/users_cycles, independence unproven
Resolution/owner: Empirical → business rule — Data + Product
Next action: Reconcile same identifiers and windows from source SQL; do not relabel false as customer failure
Source: C4 §10; C5; U1 S4

F044 · no_success

Stage: Product attempt
Field/fact: no_success
Meaning gap: Counter is not an owned failure verdict; nonzero behavior uncommissioned
Independent recording: Candidate only; independent lineage/record availability unconfirmed: raw attempts/result codes ↔ app activity counters
Resolution/owner: Empirical → business rule — Client engineering + Product
Next action: Recover source contract and test keys/windows before promotion
Source: C5 defects; C4 §3

F045 · airbyte.user_usage counters / date grain

Stage: Product attempt
Field/fact: airbyte.user_usage counters / date grain
Meaning gap: Repeated ID/date rows; snapshot/increment and future dates unresolved
Independent recording: Candidate only; independent lineage/record availability unconfirmed: raw traffic/session event totals ↔ counter rows; lineage unconfirmed
Resolution/owner: Empirical — Telemetry producer + Data
Next action: Recover producing SQL; test ANY/EVERY reductions and counter units against independent source
Source: C1 usage-source correction

F046 · avg_connection_time_from_connect_to_vpn

Stage: Product attempt
Field/fact: avg_connection_time_from_connect_to_vpn
Meaning gap: Stored average, missing event-count weight, success/error mixture, Windows pseudo-id
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source event gaps ↔ stored averages share likely lineage; not independent validation
Resolution/owner: Empirical — Telemetry producer
Next action: Recover producer SQL and event denominator; test against source-owned independent timing before performance claims
Source: C1 timing correction; C2 (superseded interpretation)

F047 · eligible attempt

Stage: Technical outcome
Field/fact: eligible attempt
Meaning gap: Which initiated attempts belong in the success denominator
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Business rule required — Product
Next action: Locate adopted eligibility contract and positive/negative examples
Source: C4 §5

F048 · success / partial / failure / unknown verdict

Stage: Technical outcome
Field/fact: success / partial / failure / unknown verdict
Meaning gap: Raw connected/traffic/payment is not technical success by definition
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client and server evidence; owned verdict definition unresolved
Resolution/owner: Business rule required; empirical implementation test — Product + engineering
Next action: Product owns verdict/exclusions; validate implementation against independent client/server observations
Source: C4 §5

F049 · disconnect / recovery

Stage: Technical outcome
Field/fact: disconnect / recovery
Meaning gap: Normal end versus failure; recovery time and resumed usability
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client recovery event ↔ server continuation
Resolution/owner: Empirical → business rule — Product + engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F050 · customer-job outcome

Stage: Technical outcome
Field/fact: customer-job outcome
Meaning gap: Customer task success cannot be inferred from traffic/payment/silence
Independent recording: Separate support and survey text exists; a linked explicit job-outcome recording is unconfirmed
Resolution/owner: Customer recording + business rule required — Product/Research
Next action: Read existing screened voice first; preserve unknown job outcome where no explicit account exists
Source: C3/C5; C4 beyond-Simonas table

F051 · evidence provenance / verdict version

Stage: Technical outcome
Field/fact: evidence provenance / verdict version
Meaning gap: Source identity, algorithm version, time and auditability
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source run/version metadata ↔ derived verdict
Resolution/owner: Empirical — Data + Product
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F052 · external_id / contact_id

Stage: Support
Field/fact: external_id / contact_id
Meaning gap: Identity coverage, contact merges and effective-dated account ownership
Independent recording: Candidate only; independent lineage/record availability unconfirmed: support source ID ↔ account/customer identity
Resolution/owner: Empirical → business rule — Support + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C3 limits; C4 §4; U3

F053 · conversation_id / conversation parts

Stage: Support
Field/fact: conversation_id / conversation parts
Meaning gap: Conversation versus support episode; participant and message-type semantics
Independent recording: Candidate only; independent lineage/record availability unconfirmed: support source conversation ↔ part records; same-system corroboration only
Resolution/owner: Empirical → business rule — Support + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F054 · incident grouping / linked attempt

Stage: Support
Field/fact: incident grouping / linked attempt
Meaning gap: One incident across conversations/devices; accepted incident/attempt edge missing
Independent recording: Candidate only; independent lineage/record availability unconfirmed: incident log ↔ support messages and attempt telemetry
Resolution/owner: Empirical → business rule — Support + Engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F055 · support timestamps / first reply time

Stage: Support
Field/fact: support timestamps / first reply time
Meaning gap: Reply-time fields disagree; business-hours, bot/human and timing basis
Independent recording: Candidate only; independent lineage/record availability unconfirmed: message-part timestamps ↔ Intercom reply-time field (same source, not independent)
Resolution/owner: Empirical → business rule — Support + Data
Next action: Recompute from event parts; obtain support-hours and qualifying-reply rule; second independent time log unconfirmed
Source: C5 defects; C4 §5

F056 · actions / escalation

Stage: Support
Field/fact: actions / escalation
Meaning gap: Task status versus action actually carried out; stable incident edge
Independent recording: Candidate only; independent lineage/record availability unconfirmed: support event ↔ ClickUp task/history
Resolution/owner: Empirical → business rule — Support + Engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F057 · closed / resolved

Stage: Support
Field/fact: closed / resolved
Meaning gap: Closed workflow state versus customer-confirmed resolution
Independent recording: Separately authored recording exists; linkage unchecked: customer replies/free text are separate authored recordings from agent closure; coverage/linkage unconfirmed
Resolution/owner: Business rule required + empirical corroboration — Support
Next action: Recover closure/resolution policy; check linked customer follow-up without treating silence as success
Source: C4 §3/5; C5

F058 · recorded labour / handle time

Stage: Support
Field/fact: recorded labour / handle time
Meaning gap: Actual work versus elapsed/reply time; activities and units
Independent recording: Candidate only; independent lineage/record availability unconfirmed: time entries ↔ support/engineering work records; neither full denominator established
Resolution/owner: Empirical → business rule — Support + Engineering
Next action: Inspect authorized retained time-entry/worklog routes before asking whether time is generated
Source: C4 §5/11; C5

F059 · App_usage_cancel_reasons.feedback

Stage: Voice
Field/fact: App_usage_cancel_reasons.feedback
Meaning gap: Free-text coding completeness and respondent/response denominator
Independent recording: Candidate only; selected reasons and free text belong to the same submission. Independent candidates: support conversation text or telemetry for the same user; linkage and lineage unconfirmed
Resolution/owner: Empirical coding + interpretation — S2 writer + Product/Research
Next action: Read 3,947 responses / 2,255 distinct texts against one codebook; retain mixed/unknown classes
Source: U1 S2

F060 · app_usage_sessions.feedback

Stage: Voice
Field/fact: app_usage_sessions.feedback
Meaning gap: STRING schema known; populated content and scope unknown
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical — S2 writer
Next action: Aggregate null/non-null/grain first, then screened sample; no raw text in shared output
Source: C3; U1 S2

F061 · app_usage_sessions.reasons

Stage: Voice
Field/fact: app_usage_sessions.reasons
Meaning gap: STRING schema known; taxonomy and null meaning unknown
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical — S2 writer
Next action: Profile values and locate producer taxonomy before treating values as customer reasons
Source: U1 S2

F062 · review prompt positive / negative / feedback-opened state

Stage: Voice
Field/fact: review prompt positive / negative / feedback-opened state
Meaning gap: Prompt event semantics, eligibility/branching and denominators; not actual store rating
Independent recording: Candidate only; independent lineage/record availability unconfirmed: client prompt event ↔ review flow definition; store aggregate not same-person cross-check
Resolution/owner: Empirical → business rule — Product + S2 writer
Next action: Map real columns/event values, count all states by period/platform; unknown exact physical field name retained
Source: C1; U1 S2

F063 · active / retained

Stage: Lifecycle
Field/fact: active / retained
Meaning gap: Meaning: usage, entitlement, payment or explicit relationship; equal windows
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source usage/payment/entitlement each records a different fact
Resolution/owner: Business rule required — Product + Finance
Next action: Choose and version the decision-relevant definition; report component observations separately
Source: C4 §5

F064 · churned / lapsed

Stage: Lifecycle
Field/fact: churned / lapsed
Meaning gap: Cancellation, expiry, failed renewal and inactivity are different clocks
Independent recording: Candidate only; independent lineage/record availability unconfirmed: lifecycle notifications ↔ entitlement/payment histories
Resolution/owner: Business rule required + empirical test — Product + Finance
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F065 · reactivated / recovered payer

Stage: Lifecycle
Field/fact: reactivated / recovered payer
Meaning gap: Gap and return definitions; same account/payer/window
Independent recording: Candidate only; independent lineage/record availability unconfirmed: renewed payment ↔ entitlement/usage return
Resolution/owner: Business rule required + empirical test — Product + Finance
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F066 · D30 / D90 / D180 observation window

Stage: Lifecycle
Field/fact: D30 / D90 / D180 observation window
Meaning gap: Day zero, maturity, censoring and missing-event treatment
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Business rule required — Decision owner + Data
Next action: Freeze day-zero and horizon; keep immature and timestamp-unknown accounts in denominator
Source: C4 §5

F067 · recognized revenue

Stage: Economics
Field/fact: recognized revenue
Meaning gap: Recognition basis differs from cash settlement
Independent recording: Candidate only; independent lineage/record availability unconfirmed: Finance ledger ↔ processor settlement (different facts)
Resolution/owner: Business rule required — Finance
Next action: Locate adopted recognition policy and component ledger; do not substitute settled cash
Source: C4 §5

F068 · affiliate fees / commission

Stage: Economics
Field/fact: affiliate fees / commission
Meaning gap: Accrual versus approved/paid, reversals and order attribution
Independent recording: Candidate only; independent lineage/record availability unconfirmed: CJ/impact source commission ↔ warehouse order/component
Resolution/owner: Empirical → business rule — Finance + Marketing
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F069 · cycle_resi_cost_usd

Stage: Economics
Field/fact: cycle_resi_cost_usd
Meaning gap: Rate-sheet model versus actual node payout; attribution window
Independent recording: Candidate only; independent lineage/record availability unconfirmed: payout statement ↔ per-cycle model; sheet and model may share input
Resolution/owner: Empirical → business rule — Finance + network owner
Next action: Recover source contract and test keys/windows before promotion
Source: C1; U1 cost bridge; C4 §5

F070 · cycle_dc_cost_usd

Stage: Economics
Field/fact: cycle_dc_cost_usd
Meaning gap: Datacenter cost components, actual settlement and allocation
Independent recording: Candidate only; independent lineage/record availability unconfirmed: provider invoice ↔ cost model
Resolution/owner: Empirical → business rule — Finance + network owner
Next action: Recover source contract and test keys/windows before promotion
Source: U1 cost bridge; C4 §5

F071 · cycle_total_gb_cost_usd

Stage: Economics
Field/fact: cycle_total_gb_cost_usd
Meaning gap: Sum formula, included components and missing terms
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source payouts/invoices ↔ model; independence unconfirmed
Resolution/owner: Empirical → business rule — Finance + Data
Next action: Recover source contract and test keys/windows before promotion
Source: U1 cost bridge; C4 §5

F072 · residential / datacenter GB fields

Stage: Economics
Field/fact: residential / datacenter GB fields
Meaning gap: Bytes versus GB/GiB, billable directions, deduplication and cycle window
Independent recording: Candidate only; independent lineage/record availability unconfirmed: network usage source ↔ per-cycle traffic; shared lineage possible
Resolution/owner: Empirical → billing rule — Network + Finance
Next action: Recover source contract and test keys/windows before promotion
Source: C1; U1 cost bridge

F073 · cycle_profit_usd

Stage: Economics
Field/fact: cycle_profit_usd
Meaning gap: Model contribution before payroll/ads/tools/infra versus full profit
Independent recording: Candidate only; independent lineage/record availability unconfirmed: independent settlement and cost components; full set not established
Resolution/owner: Business rule required + empirical reconciliation — Finance
Next action: Retain measured component contribution; commission netting and all missing components before full profit
Source: U1 correction; C4 §5; C5

F074 · support / engineering labour cost

Stage: Economics
Field/fact: support / engineering labour cost
Meaning gap: Recorded time × actual allocation rule; missing time is unknown
Independent recording: Candidate only; independent lineage/record availability unconfirmed: worklogs/payroll cost basis; retained account-level linkage unconfirmed
Resolution/owner: Business rule required + empirical time evidence — Finance + Support/Engineering
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F075 · CAC

Stage: Economics
Field/fact: CAC
Meaning gap: Acquisition spend inclusion, cohort/window attribution and organic treatment
Independent recording: Candidate only; independent lineage/record availability unconfirmed: source ad/affiliate spend ↔ joined acquisition exposure; customer denominator open
Resolution/owner: Business rule required + empirical attribution — Finance + Marketing
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F076 · attributable abuse cost

Stage: Economics
Field/fact: attributable abuse cost
Meaning gap: Loss components, causal attribution and double counting
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Business rule required + empirical case evidence — Finance + Risk/operations
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F077 · net contribution / netting rule

Stage: Economics
Field/fact: net contribution / netting rule
Meaning gap: Refund, dispute, tax, fees, delivery, labour and acquisition treatment
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Business rule required — Finance
Next action: Finance owns component inclusion/allocation; one known missing term keeps result partial
Source: C4 §5

F078 · intervention / experiment_id

Stage: Decision exposure
Field/fact: intervention / experiment_id
Meaning gap: Stable intervention version and account-level mapping
Independent recording: Candidate only; independent lineage/record availability unconfirmed: decision/task log ↔ experiment system
Resolution/owner: Empirical → business rule — Decision/experiment owner
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F079 · eligibility / assignment

Stage: Decision exposure
Field/fact: eligibility / assignment
Meaning gap: Eligible population and randomization/assignment rule
Independent recording: Candidate only; independent lineage/record availability unconfirmed: assignment log ↔ eligibility query; recording unestablished
Resolution/owner: Business rule required + empirical implementation test — Experiment owner
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F080 · account exposure

Stage: Decision exposure
Field/fact: account exposure
Meaning gap: Assigned versus actually exposed; campaign metadata is not exposure
Independent recording: No second recording established in the named evidence; source search remains open
Resolution/owner: Empirical → business rule — Experiment owner + Data
Next action: Find account-level exposure records and source event; preserve unexposed/unknown separately
Source: C4 §5

F081 · decision date / owner

Stage: Decision exposure
Field/fact: decision date / owner
Meaning gap: Adopted decision and accountable owner versus discussion/task update
Independent recording: Candidate only; independent lineage/record availability unconfirmed: approved decision record ↔ action/task history
Resolution/owner: Business rule/authority record required — Decision Steward
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F082 · mature outcome window

Stage: Decision exposure
Field/fact: mature outcome window
Meaning gap: Outcome definition, baseline, measurement date and causal comparison
Independent recording: Candidate only; independent lineage/record availability unconfirmed: pre-registered plan ↔ outcome event ledger (different roles)
Resolution/owner: Business rule required + empirical evaluation — Decision owner + Data
Next action: Recover source contract and test keys/windows before promotion
Source: C4 §5

F083 · users_cycles per-cycle activity/health: successful/failed requests, error codes, days active, sessions, seconds online, health status, first-open latency, churn flag/reason

Stage: Product attempt
Field/fact: users_cycles per-cycle activity/health: successful/failed requests, error codes, days active, sessions, seconds online, health status, first-open latency, churn flag/reason
Meaning gap: Units, aggregation windows, outcome mapping and coverage; a per-cycle summary does not itself define product success
Independent recording: Candidate: raw app_usage tables; independence and lineage relative to vpn_purchases unresolved, and excluded raw tables remain outside default analysis authority
Resolution/owner: Empirical → business rule — Data + Product
Next action: Recover producer SQL and per-cycle grain; reconcile only within existing read authority, then commission the success rule
Source: U3 addendum

Every read gap: named denominator, evidence that would overturn it, and expiry at the next matching receipt or 16 September 2026. No whole-estate exhaustion claim.

Read structured ledger data · Source: Kairos — Coverage Ledger · 44e70dd51d39

Revenue, total · July

$288k/mo

deck-reported, +9% MoM · implies June $264k — 3.5× the old ~$75.6k June baseline, now downgraded to unverified recollection (likely partial perimeter). Entity, gross/net and recognition basis all undefined · 08-07 audit perimeter open

VPN subscriptions

Recovered to May level

the −10.2%/mo melt is downgraded from trajectory to one earlier observation · July MRR not shown; actual LTV below target, gap unquantified · 08-07 insider

SuperProxy (= GoProxies? unmapped)

$11.6k July sales

136% of an $8.5k target — but 78% is paid tests ($9,084), not durable MRR ($2,544) · deck/vault product naming unreconciled · 08-07 conflict

Treasury

~€30M

denomination, owner, authority all unknown · Šaras 07-10 verbal

Cost base (the old "burn" figure, corrected)

~€4M/yr

≈€333k/mo, derived: revenue ~€2M/yr + net loss ~€2M/yr (Šaras's actual words, 07-10: "revenue is 2M, loss is 2M") · CORRECTED 08-12 — the earlier "burn ~€2M/yr" collapsed net loss into total spend and halved the cost base; €4M ÷ ~40 people ≈ €100k/head resolves the audit's implausibly-low flag · runway unchanged ~15yr (keys off net loss) · breakdown unknown corrected · verbal inputs

Term · "MN DB"

= the BigQuery warehouse

mysterium-bq, 24 datasets / 503 tables — nothing wider: not production Postgres as infrastructure (backups, DR, load: never audited, no access), not the Kairos Forgejo host. Lee, verbatim 08-16: "When I say MN DB (at least for now, until agreed on another definition) I mean BigQuery warehouse." Evidence: Kairos CONTEXT/warehouse/MANIFEST.md definition

VPN July gross (warehouse, direct)

$162.3k clean

curated claude.vpn_purchases, direct BigQuery 08-11 · the BI feed (mv_user_revenue_actual / looker) carries zero-length duplicate rows inflating the same population to $230.3k — a data-governance failure with a low-probability fraud-risk branch, private and unnamed · assessment duplication fact; clean value working tier; unit untested against a refuted sibling; intent unknown

MYST token

$0.119

cap ~$3.9M · ~$40k/day · MEXC-led (~81%; rest Quickswap/Uniswap DEX) · 7d +24.5% — up ~30% vs the 07-18 read ($0.090 / $2.9M) · listing status (no Binance; Coinbase/KuCoin pages non-trading) last verified 07-18 · 08-13 CoinGecko solid, time-sensitive

VPN churn by country tier (warehouse, direct, month-start cohort)

38% tier 3 · vs 25% tier 1 · July gross

net of reactivation within 30 days: 28% vs 19% · tier 3 ≈1.5× tier 1 every month Feb–Jul · one list price across tiers on the direct gateways ($13.49 plus-plan modal in every tier; Google Play tier 3 ≈7% lower) · ≈48% of 2026 churned subscriptions carry a billing-retry label · Primer sits under the stripe label in the spine · re-run on the estimator the opposing review named (G7, 08-17) · note · rows W6–W8 measured, working tier — not reconciled, not owner-validated (L2)

VPN cancellation reasons (in-app survey, warehouse, direct — it was never missing)

60.7% any technical · vs 35.8% price

App_usage.App_usage_cancel_reasons: multi-select + free text, 2025-06-19 → 2026-09-01, 18,419 responses on 9,060 distinct respondents (iOS 7,014 · Android 6,419 · Windows 4,847 · Mac 192) — Other 35.9% · TooExpensive 35.8% · Disconnects 19.9% · Error7040 18.2% · Usability 18.0% · Testing 17.3% · then a 12–15% tail. Peaked Feb 2026 (1,947/mo), now ~900/mo; app-surface only, self-selected. Separately postgre.mv_vpn_cancel_reasons is a stale system classifier (three buckets, sync dead since 2026-04-28, a BILLING_RETRY/BILLING RETRY label typo splitting 72k rows) — not a survey · 09-02 read solid

MN DB audit · coverage (analytical-fitness audit, plan v1.0 executed 08-16/17 — scored, never "complete")

184/204 live tables scored whole · 15.1 % of live bytes

headline, verbatim from the audit's own COVERAGE.md: scored_whole 184/204 (15.1 % of live bytes) · partial 1 (58.4 % of live bytes, partition-stratified — not scanned) · unscored bytes 26.5 % · sampled evidence: 1 table(s) (App_usage.App_usage_events, 26.5 % of live bytes) — 2 draws × 2 % blocks + column-limited whole-table facts, billed ≈21.3 GB (expected 23.5 GB · admission exposure Σ caps 411.0 GB); identity at render: current; not a score (amendment v1.2.10) · empty 18 · family 3 (analytics_354300879.events_intraday_*: shards 60, days 20260619–20260817, gaps 0; analytics_354300879.pseudonymous_users_*: shards 60, days 20260617–20260815, gaps 0; analytics_354300879.users_*: shards 60, days 20260617–20260815, gaps 0) · skipped 0 · failed 1 · not attempted 0 · queried 175 · finding-cited 32 · source findings: acknowledged 9 · unfit 7 · fit 4 (peer categories) · evidence:unscored tags 14/478 · suspect-superseded 7 live / 18 all · a partial score is a partial score, an unscored table is named, the word complete appears in no report · the unscored 26.5 % (App_usage_events, 180 GB, unpartitioned): a bounded sampled path went eleven Talos rounds 08-17; v1.2.10 was ACCEPTED and executed 11:51Z. Q-S1, a whole-table column-limited scan, found 17,077 rows dated more than a day into the future (max 2045-01-27), billed 21.3 GB ≈ $0.13. This is evidence, not a score: the table stays unscored and 26.5 % of live bytes remain unscored · results of record: Kairos repo CONTEXT/warehouse/OPEN-REPORT-v2.md (branch audit/scorecard, held back from the shared host until Lee lifts the push pause) · every number here derives from /state/coverage.json (COVERAGE.json generated 2026-08-17T120741Z · MEMBERSHIP sha 09928359f7fe… · audit branch audit/scorecard@01ce967 · pulled 2026-08-17T120812Z) derived

LT operating ring · FY2025 (MN Intelligence + MN Technologijos, filed)

−€298k on €3.10M

net loss ≈€25k/mo · ring cost base ≈€283k/mo · 25 insured staff (13 + 12), payroll ≈€113k/mo gross (~40% of ring cost) · equity −€784k carried by €1.82M liabilities — funded, not self-sustaining · diverges from the 07-10 verbal: filed LT revenue is above the verbal "~€2M" and the LT loss is far below the verbal "~€2M", so ~€1.7M of the verbal loss must sit outside the ring (BlockDev / treasury / protocol) · runway keys off treasury + group loss, not these filings · 08-16 registry, Kairos CONTEXT/finance/ (C1) · note public filing divergence open — Šaras

Nodes / IPs

~22k · 31.5k · 32.6k

three published counts, basis unresolved — vs GoProxies' 80M+ IP claim conflict

MystNodes payouts

$568k cumulative

thin per-node earnings · operator = NetSys · 20% network fee · public stats read 07-17 public

Engagement

Running, unsigned

both decision nodes approved in principle 07-31 (Ro + Gitanas; Paulius engaged) · one month, cancellable — continuation on substantial results (Lee correction, 08-07) · €50/h per member, 200h combined cap, ~€300/mo traffic valve, ceiling €10k/mo · payer = Ro's MB "attached to MN" (08-01; name/code + upside authority still open) · work started Mon 08-10 — agreement still unsigned, Robertas himself notes it in the 08-11 message · re-contract clauses drafted, Lee-gated

September board target · first-sprint review Fri 09-11

Written, 08-11 · review 09-11

First-sprint outcome review: Friday 2026-09-11, ~3 h reserved, other work cleared (accepted in principle on the 08-28 call; judges work done, results produced and where Kairos has arrived — not process sophistication; sprint 1 = 08-10 → 09-11, 33 days; attendees and continuation-decision owner unconfirmed; the 08-11 pack ask below has not been mentioned by either side since — stands or folds in: a question for the 09-04 call). Robertas's written request: a pack circulated 3 days before the board meeting (end of the first week of September) — 3–5 one-page candidate opportunities in one comparable structure, ≥1 internal + ≥1 external (or an evidenced empty side), ranked recommendation, one-line rejection log; each candidate carries its asymmetry, MN-specific advantage, cheapest prove/kill test, kill condition, and next step with cost + timeframe. Business plans, financial models, built product excluded. · Lee's scope correction: screen 3–5 · deepen the strongest 1–2 · prepare one complete torpedo contract · launch at most one, only through the explicit readiness gate — no launch promised. · Lee's reply unsent · target · verbatim

Mysterium is not one business with one lever. It is a portfolio of four systems with conflicting economics — consumer VPN, GoProxies, node/protocol infrastructure, and a mission layer — wrapped in a fragmented entity map. "Break-even" is meaningless until the board chooses the economic perimeter: commercial-operating, group-consolidated, or commercial-plus-mission. The scarce resource is still desire — MN has no forcing function, and the sprint’s job is to install one — and urgency now has a shape: one decision table, one dated board choice, one cash-quality proof.

Clocks are constraints, not demand proof: Cloudflare default-blocks agent traffic Sept 15 (scoped) · NetNut/Popa takedown (07-02) reshuffles the category — displaced demand is a hypothesis with selection risk. Cadence, settled 08-14: Fridays 13:00 Europe/Vilnius for the duration, plus the client-facing one-pager, live and Lee-approved.

Red-team verdict — adjudicated 07-18

Codex (Sol) attacked the corpus zero-base; Claude adjudicated per the standing red-team lane. Where passes conflict, this verdict wins. Full reasoning: audit, final section.

Survives — act now

Portfolio reframe: perimeter decision first
GoProxies supply gate: 80M+ IPs ≠ ~22k–31.5k nodes
All calibration downgrades (n=1, verbal financials, clocks ≠ demand)
Protective contract terms → clause set
Research stop: everything further is closing ammo or paid work

Gated — internal data decides

Entity-scoped P&L + intercompany flows
12-month VPN cohort/churn series (the pivotal fork)
GoProxies concentration + contribution by SKU
Burn map with decision rights
The reseller fork: if GoProxies mostly resells upstream supply, "scale it" is a standalone margin business, not network monetization

Rejected — with reasons

Walk-away ultimatum → gate the work, not the relationship; data room = first work product
"Sol changed everything" → net-new = portfolio frame + supply gate; rest is calibration
"Kill no-KYC" → dead as promised torpedo, alive as discovery hypothesis
Schema-tag finding "refuted" → publisher tag ≠ ToS party; both data points stand

Insight ledger

Append-only, 68 entries deep. Hard-won clarity never gets washed away: dead ideas are struck, not deleted. RED-TEAMED / CORRECTED = downgraded, reframed or superseded — the pass date rides inline. Click any pointer to open the full source page in this browser.

Status key — ALIVE — standing, load-bearingDEAD / STRUCK — killed; kept visible so it isn’t re-litigatedDOWNGRADED / CAVEATED — survives with reduced forceCORRECTED / REFRAMED — superseded by the entry it namesGATED — waits on named evidenceREOPENED — a closed question, reopenedREGISTERED — logged, pending confirmationPROCESS — a working rule, not a findingVERDICT — an adjudication record

Money is UP the stack. Raw GB is commodity; the 10–30× markup lives in managed / per-success access. Reinforced 07-18: GoProxies Scraping API is in private beta. prep07-03 · ALIVE

FWA / consented-mobile supply thesis. Killed by carrier ToS, no proven trust premium, Bright Data's retreat. Survives only as a cheap node-classification test.07-03 · DEAD

Agent traffic threads the needle. Low-volume, human-paced traffic dodges carrier ToS and throttling; a decentralized trickle self-selects the premium use case. Unvalidated.07-03 · ALIVE

Ro cannot name his revenue split — the runway is the anesthetic. RED-TEAMED: an interview probe, not a diagnosis of a person or governance system. One recall failure in one meeting is not proof. frameworks07-05 · REFRAMED

The one uncontested wedge: permissionless + consented, crypto-paid, no-KYC. RED-TEAMED: a demand hypothesis with adverse-selection, abuse, procurement, and operator-risk exposure — not a trial promise. Discovery only (T1). market-map07-06 · DOWNGRADED

Token revival base rate ≈ low single digits. Every precedent needed a bigger trough, a war chest, and an external wave. Second-order amplifier only — never the plan. precedents07-06 · ALIVE

Tesonet argues both ways. The market pays €1B from this city — AND the neighbors are 100–300× better capitalized. Never ship only the flattering half.07-06 · ALIVE

The founding premise is the baggage (Dunford Step 1). Positioned as its 2016 self (dVPN/privacy) while materially a 2026 data-supply network.07-13 · ALIVE

Four waves, mistimed. Funded by the fast waves (ICO + ETH), needing the slow ones (privacy, DePIN) that crested after momentum died. Forward filter: pair one fast wave with one slow-structural. wave-map07-17 · ALIVE

MN is a VPN subscription business. The 3.21 PB is a vanity metric — never per-GB revenue math.07-17 · ALIVE

Capacity is not the constraint — demand is. RED-TEAMED: activity is proven; monetizable proxy capacity and the 5–8% ratio are not. Testable hypothesis only — sellable supply depends on geography, consent, IP class, quality, coexistence. infra07-17 · DOWNGRADED

The melting base outruns growth. A VPN base halving every ~7 months beats any year-one proxy ramp → cutting burn is the floor-setter. RED-TEAMED: decline is n=1; map costs and capability dependencies before deciding any cut is safe. plan07-17 · CAVEATED

GoProxies is in-house — the master key, answered. RED-TEAMED: public ToS names no contracting entity; ownership and intercompany economics stay open. Adjudication: the schema.org tag naming MN Intelligence UAB stands as evidence — publisher tag ≠ ToS party; both data points feed one open question. Ask: who invoices GoProxies customers?07-17 · REOPENED

Node economics are structurally contradictory. Only maximally-exposed residential operators earn; you cannot opsec out of it. node-econ07-17 · ALIVE

The origin story is a cage. Ideology caps monetization; the profitable truth is orthogonal to the founding ideology. Proton is the structural (not promised) precedent.07-17 · ALIVE

Hermes is not a payments goldmine. The market standardized on x402/L402 + stablecoins — ride those rails on top of MN access; the asset is 4.5 years of know-how.07-17 · ALIVE

Treasury ⟷ token-cap inversion; break-even may be bought, not grown. RED-TEAMED: capital allocation is an option only after treasury ownership, liquidity, authorization, and the real operating perimeter are verified. M&A is never Phase 1 — buying revenue before proving profit and clean supply can buy liability. audit07-18 · GATED

We sell urgency, not analysis. MN has no forcing function; the trial needs adoption gates and dated internal decisions. RED-TEAMED: never use external clocks as buyer-demand proof.07-18 · ALIVE

Narrative problem = fragmentation, not silence. The one consistent engine (VPN news desk, near-daily) aims at the declining product. sweep07-18 · ALIVE

Entity-scoped break-even: the revenue entity is ~€22k/mo from break-even. RED-TEAMED — invalid as stated: FY2025 net loss is not current monthly burn; BlockDev's burn ownership is unproven. Entity mapping stays an open gate, not a conclusion.07-18 · STRUCK

Consent is the enforcement line (NetNut/Popa). RED-TEAMED: consent evidence is a counsel-reviewed operating gate, not a legal or commercial moat; the displacement window is a hypothesis with selection risk — displaced buyers may be exactly the ones to refuse.07-18 · DOWNGRADED

An abandoned blog is negative, not zero. Commit to a predefined cadence or don't start; or run a time-boxed high-quality test, then keep-or-kill. Proven by MN's own portfolio.07-18 · ALIVE

Negative findings carry their scope. "Not found on X, as of DATE" — or the claim doesn't travel. Corollary from the schema-tag save: a finding also carries its evidence type; disagreement is only refutation when two sources measure the same thing.07-18 · PROCESS

Red-team verdict: Phase 1 survives; the supporting certainty does not. 3.21 PB, 5–8%, €30M runway, €22k-to-break-even, and the crypto-native wedge are hypotheses, not facts, until the paid data pulls land. audit07-18 · ALIVE

Mysterium is a portfolio, not one business. Commercial VPN, GoProxies, node/protocol infrastructure, and the mission layer need separate economic boundaries. MYST holders and node runners are constituencies, not revenue lines.07-18 · ALIVE

GoProxies source truth is a gate. 80M+ advertised IPs across residential/ISP/datacenter SKUs cannot be equated with MN's ~22k–31.5k nodes. Map source, transfer price, consent, quality, and contribution margin per SKU before "scaling the proxy motion." Opens the reseller fork.07-18 · ALIVE

Fragmentation may be a firebreak, not merely a marketing failure. Public surfaces mix BlockDev, NetSys, MN Intelligence, and an unnamed GoProxies counterparty. Resolve legal purpose before consolidating brands or trust claims.07-18 · ALIVE

No-KYC is adverse selection, not a moat. It may attract the exact abuse and procurement risk that destabilises node supply. Adjudicated: dead as a promised torpedo, alive as a discovery hypothesis (T1 stands).07-18 · ALIVE

The engagement itself needs a decision gate. Adjudicated: pre-signature gates shrink to entity/signer + payment protection + no-attribution; the data room becomes the sprint's first work product. Gate the work, not the relationship.07-18 · ALIVE

Claude adjudication of the Sol pass — final verdict. Survives: portfolio frame, supply gate, all downgrades, protective terms, research stop. Gated: the four data pulls + the reseller fork. Rejected: walk-away ultimatum; "Sol changed everything" framing; "kill no-KYC" over-correction. audit07-18 · VERDICT

The money person is the floor, not the ask (Lee's correction) — one named 48h money-owner is mandatory but covers a sliver; the rest is distributed and written down nowhere. Access design: listening access — Slack presence + opt-in coffees, zero meetings added, framed as listening, never audit interviews.07-23 · ALIVE

Commercial terms corrected by Šaras — €50/hour per team member pre-tax; 200 combined hours/month across both unless Robertas approves another cap. Supersedes the fixed-retainer wording. A supplied signature is a proposal visual, not an executed agreement.07-24 · ALIVE

Verbal green light with a floated three-month pre-commitment. CORRECTED 08-07 by Lee: the commitment is one month, cancellable; continuation depends on substantial results. Remaining contract gates: entity, signer, payment mechanics, owners, data access, the Mammoth Framework.07-24 · CORRECTED

People-performance is an explicit workstream, not an unsayable side effect — Ro confirmed one remote employee moonlighting and said co-founder Valdas Petrulis will be let go (separate facts — do not conflate). Evidence-gate every person-level conclusion.07-24 · ALIVE

Authority map has a new candidate node — Gitanas Kancerevyčius, Swiss "mothership" director/CEO (reported), likely decision-maker alongside Ro. Establish the actual decision-rights map before promoting titles to fact.07-24 · ALIVE

Kairos has a separate funding ring — paid from the MN fund/"mothership," outside Ieva's operating entities. Protects her ring; creates a governance seam: sponsor, payer, signer, budget owner and implementer can now be different people.07-24 · ALIVE

High-quality MN data may become queryable through Claude — Šaras: Simonas will expose operational data "via Claude" soon. Would collapse the data-room bottleneck; not access yet — source tables, definitions, read path, permissions, audit trail all unknown.07-24 · ALIVE

Agree the mammoth protocol before the hunt — Ro explicitly wants upside-sharing written before a major opportunity is found (prior team fractured after success). Lee agrees: no greed, ever — but no slop either. Keep the register + attribution ledger; sign the instrument when the class turns concrete.07-24 · ALIVE

"Most likely has some excel sheets" is the finding — the funders can only guess whether marketing's measurement layer exists. Two gaps: visibility (cheap) and no reference standard (the real one). Generalises: ask of any MN function whether a standard exists before asking whether it performs.07-31 · ALIVE

The best openings arrive unprompted from inside — the marketing gap was raised by Paulius, a technical co-founder, framed as agent-automated marketing. A principal-surfaced opening carries built-in consent. Candidate Ring-1 proof shape; people-lane rule holds.07-31 · ALIVE

Preparation held under adversarial reading — Gitanas probed all call; Lee had a prepared answer each time. First live test of the corpus against a hostile-by-design reader: passed. His challenge content is worth more than the fact it was answered — partially recovered as insight 49.07-31 · ALIVE

The missing measurement layer is not a marketing problem — if no central place exists where operational data lands, every function is measured only by its owner and none can be compared to a standard. Marketing is where it became visible, not where it lives. Explains 39; test cheaply before building on it.07-31 · ALIVE

Choose the three-person substrate as if thirty will use it — staged rollout contains one silent failure: the pilot substrate never gets re-decided when people are added. Git history, atomic commits, revocation, backup look like ceremony at three and are the product at thirty.07-31 · ALIVE

Lee's credential is demonstrable — and that is also the trap — he operates a running second brain, which almost no advisor can show. The pull to resist: MN's constraints are not Lee's. Transfer the architecture and failure modes, never the implementation.07-31 · ALIVE

MAMMOTH GATE FIRED — org-wide MN intelligence layer registered — an org intelligence layer is not a €50/hour deliverable; it is mammoth-class, and Ro already argued the sharing protocol precedes the hunt. State: qualifying — pending Lee's confirmation and MN's written acknowledgement.07-31 · REGISTERED

Approval arrived before the price, the rules, and the ask. CORRECTED 08-03: Ro already knew cards 14–17 from the earlier proposal call ("buhalterija" — routine). The surviving narrower version: the undelivered cards were new to the room's two cold readers — whether Gitanas and Paulius know the ceiling and upside architecture is genuinely open. Rule: "not delivered in this room" ≠ "unknown to the counterparty."07-31 · CORRECTED

A structural decision was made in passing — Ro offered an MB "attached to MN" on a phone call; Lee said "works for me." Payer side is now clear; convert to a named entity with a defined role, settled together with the upside card, not after.08-01 · ALIVE

Ro spends comfort to protect momentum — declined a softer Monday slot so the 08-10 start would hold. A sponsor who wants this moving. The reciprocal obligation: the weekly one-pager cannot slip; continuation rides on substantial results.07-31 · ALIVE

Gitanas's doubt #1 — "don't just report the problem" — his fear: a duo that ends each month saying "revenue low, fix it." The answer that landed: analysis → hypothesis → torpedo → data. Unnamed tension: autonomy biases toward torpedoes that don't touch product — the analysis-only trap. Resolution: classify every torpedo by product-surface contact and pre-agree the insider path.07-31, recovered 08-03 · ALIVE

The objection produced the opportunity — and the system was the argument — Gitanas challenged AI slop; Lee showed his actual system; Gitanas then said "maybe we should have it company-wide." Two teeth: any generic Kairos artifact is now the exact promised failure; and the vault-copy pull is a live trigger — transfer architecture, never implementation.07-31 · ALIVE

The deliverable is the agenda — the four undelivered decisions were built into week 0 of the one-pager, so the call is led from a page the client sees. Rule: when a recurring deliverable and an agenda want the same content, ship the deliverable and lead from it.08-03 · ALIVE

A guard that ships an unenforceable rule is honest about it — the one-pager generator refuses five defects, each seen failing before trusted; the two it cannot enforce are written beside the gates. The red-first pass immediately exposed a parser bug that had masked two gates.08-03 · PROCESS

One month, then proof earns continuation — Lee corrected the carried three-month assumption: initial commitment is one month, cancellable. Kills the three-month exposure total as a current term; the €10k/mo ceiling stands; the agreement remains unsigned.08-07 · ALIVE

One all-hands deck found a contradiction; the archive can test the operating system — Lee will request the authorized historical All Hands archive. Target: longitudinal — definition drift, bet → outcome, forecast calibration — not deck summaries. Slides are management narrative, not audited truth. audit08-09 · ALIVE

The September output is a ranked decision portfolio, not a research archive. Robertas supplied the acceptance contract: 3–5 comparable one-page candidates, rejection log, recommendation. CORRECTED 08-11 by 56 + 58: the inventory exists, but "test" was still being used too loosely. target08-11 · CORRECTED

Candidate count is not the September constraint; signal strength is. Lee challenged the unconditional acceptance; recommend only what clears an explicit evidence threshold, even if only one survives. CORRECTED 08-11 by 58: screening 3–5 is credible; calling them tested is not.08-11 · CORRECTED

Best evidence and best asymmetry are currently different candidates — Paid-Test Conversion leads on evidence; Payment-Recovery is the close-to-cash challenger; Residential Vantage is the strongest external hypothesis. CORRECTED 08-11 by 58: validation priorities, not launch-ready opportunities — the names "Strike" and "Torpedo" outran readiness. hypotheses08-11 · CORRECTED

Screening, validation, launch and measured result are four different stages — Lee caught the semantic laundering in "seriously test 3–5." The honest four-week contract: screen 3–5 · deepen the strongest 1–2 to decision-grade evidence · prepare one complete torpedo contract · launch at most one, only if the readiness gate (data, named MN owner, stable baseline, reversible intervention, pre-registered metrics, observation time) clears early enough. A no-launch result can still be the ranked portfolio. MN-side access friction is a delivery dependency, not hidden inside Kairos's result obligation. target08-11 · ALIVE

A material warehouse anomaly is an audit branch, not a personnel conclusion — direct BigQuery found a duplicate-carrying BI revenue feed and a clean curated comparator. The pattern fits an ETL/SQL defect far better than fraud; intent cannot be inferred from duplicated rows. Preserve evidence, trace source SQL and consumers, reconcile externally — ask who knew what only after lineage is established. Private and unnamed. assessment08-11 · ALIVE

Upside exists in architecture, not in force — and the leverage window is closing — live terms are €10k/month fiat with no upside instrument: if Kairos generated €3M tomorrow it would capture €0 of it. The Mammoth Protocol already specifies the right shape — a time-limited share of incremental gross profit against a pre-registered baseline — so MN’s net loss never touches Kairos compensation. Delta-based pricing needs no P&L, which decouples the two asks: the gap number is needed for the mandate, never to price Kairos. Measurement rights are the enforcement mechanism ("without measurement rights, any percentage is decorative"): every warehouse query still runs through Šaras’s principal, informally — which reclassifies the Lee-owned BigQuery grant from plumbing to commercial protection. And leverage moves one way: negotiating a share after showing MN the winner is negotiating face-up. Today there are candidates and no winner — maximum leverage, closing toward the early-September board date. Clause 10 already says the upside agreement precedes the hunt; the hunt is running. upside08-12 · ALIVE

Unscouted territory is ranked by value of information, not by evidence — the blindspot map turns ten unknowns into quests across warehouse, human, and corpus lanes. Each quest names its prior, unlock, owner column, pre-written decision rule, predicted time to signal, and drop-by date. Its event record is append-only; its kill test is explicit: if no session claims a row by day 21, the map failed as a tool. Metric lineage and the owner map lead the human lane because they are the shortest path from an L2 claim to decision-grade evidence. quest map08-16 · ALIVE

A profile is not a fitness contract — and the red team ran before the board, not after — broad warehouse capture while the borrowed token lived was the right use of the hours; use-specific claims from tables whose grain, identity, date field and joins had never been contracted were not, and produced the withdrawn headlines. The opposing review landed at 23:52 and every test it named that the warehouse could answer ran before anything reached the board: the entitlement-cohort test killed an iOS first-connect row that would otherwise have shipped (entitled first-tries connect 96.9% / 96.3%; the gap was composition), the month-start cohort reproduced the tier ratio (tier 3 ≈1.5× tier 1), within-platform standardisation kept the survey direction and stripped the causal claim, an identical-population rerun replaced "no revenue missing" with "no aggregate discontinuity within 2%". Six critical tables now carry written contracts; a claim outside a contract is working-tier until the contract is extended. Board effect: v0.6.0 — W6 · W7 · W8 signalled, W3 reworded, retention #02 gains its tier cut. note08-17 · PROCESS

A red team of the audit’s conclusions found the render, not the warehouse, was the weak side — and the surfaces moved the same day — a sealed Claude-side (Tris) review of the eight load-bearing conclusions plus the joins finding, checked against the audit’s own artifacts and against this room, the board and the canon note, returned ACCEPT WITH MATERIAL CORRECTIONS: eight findings, all accepted by the writer; Codex holds one rebuttal on the four touching its artifacts. Six of eight conclusions stood on their evidence — the tier-churn cohort estimator, W3, W6 and the label-not-outcome discipline stand as written. What failed was wording that outran its evidence: this room said the sampled path was “not executed” beside a card that said it was; the board worded a session share as a revenue share with no untested; the 93 % conditional was stated inverted; the dollar-share label was unsupported by the safe table; W8’s “100–102 %” dropped the 110 % week; the $162.3k stamp claimed more than the canon did. Every string is now the adjudicated sentence; the board’s changelog records the correction as v0.6.1. Rule carried, not yet written: what an evidence:unscored tag obliges a surface to say. v0.6.1 changelog08-17 · CORRECTED

One sensitive field must not silently become a whole-system access veto — in Lee’s post-call account (user-reported; the transcript was still finalising), employee compensation was raised as private data and treated as a blocker to broader access. The narrower requirement Lee proposed: establish the lawful, contractual and authority perimeter field by field; exclude, aggregate, mask, pseudonymise or purpose-limit what needs protection; open the rest through governed read access; revisit restricted fields only if the accessible perimeter proves insufficient. His “98 %” is a hunch, not coverage evidence. Agents prepare the source-backed issue map and controls; Lithuanian/EU counsel remains the legal authority, and no agent conclusion widens access. debrief08-21 · USER-REPORTED

“Agent-friendly” is an acceptance contract, not a file extension — Robertas reportedly asked that Kairos outputs be usable by agents now and later. Lee said the War Room was machine-readable, then correctly reopened whether that holds surface-wide; current status: unverified. Inventory each relevant surface before choosing architecture — human and machine interfaces, stable IDs, typed/versioned schema, provenance, freshness, access, deterministic retrieval, change semantics, explicit unknown/error states, fixtures and conformance tests. The carrier may be Markdown, SQL, JSON, API or HTML; the test is whether an authorized agent can discover, interpret, verify and safely act without guessing. First concrete fixture (09-02): given authorized access, an agent must answer “what changed on ClickUp project X since last week?” by resolving the project and window, retrieving prior and current records, citing source events and timestamps, and distinguishing no change from missing access, missing history or retrieval failure — guessing fails it. Reopen Garry Tan’s retained source before any Markdown-versus-database decision. debrief · analysis · transcript08-21 · SETTLED 09-02

The reason data was never missing — only unread — MN has run in-app cancellation-reason capture for fourteen and a half months across four platforms (App_usage.App_usage_cancel_reasons). Two exhaustive object ledgers found the table and dispositioned it correctly for their own decisions (PARKED for Payment Recovery, DECISION_CANDIDATE_CURRENTLY_BLOCKED for Retention Economics); both statuses read as “handled” on a board while one meant “nobody has looked yet”. Kairos was one gcloud auth login from the evidence and spent the week planning access asks instead. read09-02 · DIRECT BIGQUERY

Task completion is a real but SECONDARY cancellation signal — corrected the same day — on the correctly exploded multi-select (9,060 distinct respondents; reasons overlap) cancelTesting is 17.3%, ranked sixth, inside a tight 12–20% band with disconnects, Error7040, usability, missing features, speed, blocked sites, downtimes and latency; only 12.6% cite testing without also citing a technical or price reason. Leading tiers: cancelOther 35.9% · cancelTooExpensive 35.8%; any technical reason 60.7% against price at 35.8%. An earlier draft called task completion the “dominant” theme — built on a single-reason table and a free-text bucket 2nd of nine with 67% of the pool unclassified; it was wrong and is retracted. The episodic-user hypothesis stays live and unquantified against the 63,242 mature monthly cohort; it is not a headline. read09-02 · DIRECT BIGQUERY · SELF-CORRECTED

The picked list and the free text disagree about error codes — the survey option is cancelError7040 while users independently type error 7042 (14 distinct users), error 4102 (7) and error 500 (5). Either the label is wrong or distinct faults are being collapsed into one bucket. A stated-reason taxonomy that cannot name the fault correctly will under-report reliability as a cause of loss. read09-02 · DIRECT BIGQUERY

Hypothesis board

What we believe but have not proven. Nothing here is a fact yet.

Hypothesis	Status	Next test
MN has no central operational data layer	OPEN — one cheap question falsifies	Ask where an MN operator finds last month's numbers for a function they don't own. Load-bearing for the marketing standard gap and the org-hub opportunity (insights 42–43)
Commercial vs mission boundary	OPEN — board-level	Perimeter definition is the sprint's first deliverable; break-even is meaningless without it — the 08-07 all-hands $288k vs old $75.6k gap is this question made visible
VPN decline: fixable vs secular	OPEN — sharpened 08-07	All-hands: subscriptions recovered to May level; the −10.2%/mo melt downgraded to one earlier observation. Still the pivotal fork — the 12-month cohort/churn series remains the first paid pull
GoProxies supply + risk-adjusted unit economics	OPEN — Phase-1 gate	Revenue by SKU × source × consent × cohort × contribution margin; answers the reseller fork. 08-07 wrinkle: SuperProxy $11.6k July is 78% paid tests, not durable MRR
T1 · Crypto-native no-KYC wedge	discovery only	Named-buyer discovery; never a promised torpedo
T2 · Pricing / margin probe	armed	GoProxies public sheet in hand; wholesale-vs-retail gap is the first pull
T3 · Trust-surface sprint	HOLD	No external trust claim until source controls, screening, incident handling, and counsel-backed claims exist
M&A · Buy proxy revenue with treasury	backlog, gated	Only after treasury ownership/liquidity/authority + supply quality + assignability verified
Ideology-is-the-ceiling	load-bearing, unverified	Scanner watches; interview probes in the sprint
182-country breadth advantage	contested	Resolve node-stat counting basis (W1)
Paid-Test Conversion	VALIDATION LEAD 08-11	Strongest current company-reported signal ($9,084 of July's $11.6k is paid tests); economics and launch readiness unknown — reconcile the paid cohort before choosing any intervention
Payment-Recovery Candidate	CONDITIONAL EXPERIMENT 08-11	Potentially close to revenue, but recoverable value, owner, implementation and observation window unverified — do not call it a torpedo yet
Residential Vantage Network	DISCOVERY ONLY 08-11	Potentially asymmetric asset inversion; no MN demand signal, real incumbency (RIPE Atlas, Catchpoint, ThousandEyes, Kentik) — outreach authority + feasibility only, no September product test
Lawful data-access perimeter	OPEN 08-21	Lee requirement, not a legal conclusion — research Lithuanian/EU law, contract/confidentiality and technical controls field by field; separate counsel authority from MN policy; test whether masking, aggregation, pseudonymisation or purpose limitation unlocks the non-sensitive perimeter
Agent-readable Kairos substrate	OPEN 08-21 · fixture 09-02	Current War Room coverage unverified. First acceptance fixture: an authorized agent must answer “what changed on ClickUp project X since last week?” from source events and timestamps, distinguishing no change from unavailable evidence and never guessing. Inventory every relevant surface and its machine contract before recommending a future architecture or claiming the whole room is machine-readable
Churned users with no recorded connection	HYPOTHESIS 09-02 — excluded from outputs	56.7% of 119,954 churned users carry `was_connected_total = false` (~68k) in `App_usage.users_churn_reasons`; 8.5% refunded. The field’s semantics are uncommissioned — it may mean “no telemetry recorded”, not “never connected”. The 09-02 evening probe weakened it further: 94.8% of the flagged are also `opened_app_total=false`, the never-opened share is 64.2% even on Apple and 52.8% on Google where telemetry exists, and 7,594 flagged users have sessions in `users_cycles` — a user-id ↔ pseudo-id mapping gap in the table build, not behaviour. Field excluded from every output until MN supplies its definition (Simonas). read
Marketing visibility gap = missing measurement + external standard	PRIVATE — core tier only	Inference, never a personnel conclusion; moved out of the shared repo 08-11. Validate with the owner + sponsor; test whether agreed metrics plus an external benchmark change a bounded decision

Strategy state

Sol re-run, adjudicated 07-18: don't sell a 90-day growth promise — sell a Decision & Economics Sprint that earns the right to one reversible proof.

Phase 1 — the 30-day Decision & Economics Sprint

0 · Perimeter

Define commercial vs mission boundary; reconcile group/product P&Ls, intercompany flows, treasury and burn definitions.

1–7 · The decision table

Product/SKU × entity × cohort × supply source × revenue × direct cost × contribution × exposure × owner. Making this table true is the sprint’s core deliverable.

8 · The board choice

Ro + the revenue authority choose, in writing, per line: defend · harvest · grow · partner · ring-fence as mission · kill.

9–30 · One proof

One installed-base cash-quality test. Retained MRR or realised contribution margin counts; meetings, clicks, and theoretical capacity do not.

Engagement state (current, 08-14): approved in principle by both decision nodes 07-31; one month, cancellable — continuation on substantial results (Lee, 08-07); work running since Mon 08-10 — day 5 of 30, 25 left, agreement still unsigned. Cadence settled 08-14: Fridays 13:00 Europe/Vilnius, standing for the duration of the engagement — the last open item from the 07-31 room's calendar, closed; the client-facing one-pager at kairos.leematulis.com/weekly ships with each call, drafted Thursday. The month now has a written finish line: Robertas's 08-11 September board target (pack circulated 3 days before the early-September board meeting) under Lee's scope correction — screen 3–5 · deepen 1–2 · one torpedo contract · launch at most one through the readiness gate; Lee's reply unsent. Standing agenda: the four decisions the 07-31 room never reached (our €10k/mo ceiling · upside rules before the hunt · the four asks) — new to Gitanas and Paulius, already known to Ro. Torpedo rule from Gitanas's doubt #1: classify every torpedo by product-surface contact and pre-agree the insider path before torpedo #1.

Adjudicated engagement position (07-18, still governing): collaborative tone, not procurement. Three pre-signature gates only — named entity/signer, date-fixed payment, mutual no-attribution. The data room is the sprint's first work product, not an upfront demand. Clause set awaiting Lee's review: Re-contract Clauses. The send is Lee's.

Phase 2 — scale only the proof (held)

CZ-1 agent access × agent payments — hypothesis only. CZ-2 verified-human / provenance — highest coherence, heaviest build, hypothesis. WC-2 grant-funded censorship-resistance (OTF pays MN to be what it already is) — sleeper, non-dilutive.

Standard of performance — the shape the second half takes (Lee, 08-14)

Successor to the business-health page, which Lee pulled off the front door as a very lame initial version — the page still exists, it is simply not on a door until it earns the place. The rename is not cosmetic: business health describes a state you observe, standard of performance describes a bar you measure against and then close on. Mechanism, three steps then repeat — map and log every vital MN operation precisely · assemble a granular insider account of what world-class looks like for that specific operation · measure current against it, take the single lowest-hanging 80/20 gap, and close it once and for all. The constraint that carries the whole idea sits in step three: what gets implemented is never a one-off action item but a tiny habit that does not revert, or machine automation wherever automation is possible. An improvement that decays is a project, not an improvement. Gated on read-only data grants (ClickUp, Drive, Slack, Intercom among them — the data-access board, which replaced the narrower four-grant ask on 08-14 and carries the 21 August date): the warehouse records what happened, those sources carry why, and operations cannot be mapped precisely from the warehouse alone. Open, and first: which operations count as vital — drawing that list is the first real act of step one. Full mechanism: Standard of Performance.

Kill-list (checked and rejected — do not resurrect without new evidence)

Head-on B2B proxy sales vs Bright Data / Oxylabs (397 people, same city)
Token-first revival (base rate ≈ low single digits)
Generic AI-agent access outbound (vendor-dense, freshly VC-funded)
Consumer dVPN growth race (Nord's game, in Nord's town)
"Ethical supply" as a legal moat (Bright Data won in court without it)
FWA / consented-mobile supply thesis (carrier ToS guts it)
M&A as a Phase-1 lever (buying revenue can buy liability)
No-KYC as a promised trial torpedo (adverse selection)

Company — entities, people, products

The entity map is still incomplete — treat every ownership and P&L claim as open until the group and intercompany map is documented.

Entity	Who	Facts
BlockDev AG (Zug)	Robertas — president; Gitanas Kancerevycius ("Gitas") — board	Protocol developer, public-site operator · CHF 100k capital · authority over other group entities unproven · CHE-464.393.106
UAB MN Intelligence (Vilnius)	Ieva Matulaitienė — director; publicly billed "Mysterium CEO"	Standard VPN counterparty · registered ~2024-12 · FY2025: €1.13M revenue, −€268k, equity −€267k, liabilities €821k · insured staff 1 → 10 in Jan–Feb 2025, 13 (2026-07), avg gross salary €3.1k · GoProxies schema tag names it; GoProxies ToS names nobody — tie open · 07-18 · 08-16 registry (public filing)
UAB MN Technologijos (Vilnius)	Ieva Matulaitienė — director	Older dev entity (2018) · revenue 2021→2025: €0.60M / €0.79M / €1.43M / €2.31M / €1.97M · net +€24k / +€76k / +€8k / −€672k (2024) / −€30k (2025) · equity −€516k, liabilities €1.00M (2025) · insured staff 19 (2025-05) → 12 (2026-07), avg gross salary €6.4k · 07-18 · 08-16 registry (public filing)
LT ring, both entities · FY2025	Ieva's operating ring	Revenue €3.10M · net loss €298k (≈€25k/mo) · ring cost base ≈€283k/mo · headcount 25, payroll ≈€113k/mo gross (~40% of ring cost) · equity −€784k on €1.82M liabilities — funded, not self-sustaining · diverges from the 07-10 verbal ("revenue 2M, loss 2M"): ~€1.7M of the verbal loss must sit outside the ring · runway = treasury + group loss (ask 8: cash + committed funding, one line) · 08-16 registry (C1) · note
NetSys Inc. (Panama)	—	MystNodes contractual operator · Mysterium Dark publisher · Panama-law terms shift node-risk exposure to operators

Dual-CEO shape: Ro = protocol / treasury / mission; Ieva = the operating companies. Which P&L "needs saving" is an open question, not a conclusion — the proposal still addresses both. Fragmentation may be a deliberate firebreak between legally distinct products; resolve before consolidating anything.

Governance & people — as the 07-24 and 07-31 calls left it

Person	Role read	State
Robertas (Ro)	Sponsor; protocol / treasury / mission	Approves; calls the commercial layer "buhalterija" (routine); declined a softer slot to protect the 08-10 start; wants the upside protocol written before the hunt
Gitanas Kancerevyčius	Swiss "mothership" director/CEO (reported); likely decision node	Approved in principle 07-31 (nodded to Ro's read-out). Standing challenge: no AI slop, no analysis-only reporting. Floated the company-wide hub after Lee showed his system. Whether he knows the ceiling + upside architecture: open
Paulius Mozuras	Technical co-founder	Engaged 07-31 — chimed in repeatedly, raised marketing unprompted (agent-automated framing). Formal mandate, vote, signing authority: unknown
Ieva Matulaitienė	Runs the operating ring (MN Intelligence + Technologijos)	Kairos is funded outside her ring, from the fund/mothership (07-24) — she stays an operating/data counterpart, not the budget gate
Indrė [surname unknown]	Head of marketing	Team understaffed; measurement layer guessed at, not stated ("some excel sheets") — the standard gap, insight 39
Simonas	MN data person	Data-access call held 08-10; direct BigQuery warehouse access working since 08-11 — via Šaras's principal, informally, no post-engagement continuity (a Lee-owned grant is the open ask, reclassified 08-12 from plumbing to commercial protection). Securing it cost ~a working day — access friction is an MN delivery dependency
Valdas Petrulis	Original co-founder	To be let go (Ro, 07-24 — reported, private)

Team-size reads conflict: ~26–28 on LinkedIn vs "~40" said at all-hands; hiring frozen (Workable empty). Contracting: Ro's MB "attached to MN" is the payer; its name, code, legal meaning of "attached," and which entity can grant upside are the open pre-work items.

Products

Product	Sells	Price	Signal (07-18)
GoProxies — rotating residential + ISP + datacenter	Proxy access by SKU, self-serve checkout (card + crypto)	$2.40/GB → $1.40/GB @1TB	80M+ IPs advertised vs MN's ~22k–31.5k nodes — source mix unproven; the reseller fork
GoProxies — Scraping API	Up-stack per-success access	Private beta, unpriced	The up-stack move, already in motion
MysteriumVPN (+ Mysterium Dark)	Consumer dVPN, residential exits	$10.49–18.49/mo; 1-yr from $3.69/mo	iOS 4.7/5 across 5.1K ratings · shipping actively · i2Coalition member (Feb 2026)
MystNodes	Node-runner earn side	20% network fee	$568k cumulative payouts (public stats read 07-17) — thin per-node earnings · operator = NetSys
Necto Labs	$3M builder fund (2023)	—	Dormant-not-dead; zero portfolio news 2025–26

Content engines (the fragmentation picture)

Engine	State	Verdict
mysteriumvpn.com/blog/news ↗	Near-daily through Jul 2026 — privacy-news desk + original research	Strongest engine — consistent, aimed at the declining product
goproxies.com/blog ↗	Irregular, high-value — Censorship Index 2026 earned press pickup	Compounding already working; under-fed
mystnodes.com/blog ↗	1 product post 2026 + a 2025 SEO burst	Inconsistent — one product post in 2026 after a 2025 SEO burst
mysterium.network/blog ↗ · blog.mystnodes ↗ · Medium ↗	Dead (2023 / 2024)	Dead since 2023/24 — broadcasts decay to every diligence pass

Correction 07-18: the mothership DOES link its product doors today. The problem is inconsistent entity, inventory, and trust information across surfaces — not an absent buyer path.

Market & adjacent markets

Residential proxy / web-data supply, the agent-infra layer above it, and the DePIN comparables.

Player	Scale	Position vs MN
Bright Data	~$300M ARR → $400M target; 14 of top-20 LLM labs	Category king; court-tested (Meta, X wins 2024); origin = Hola botnet resale
Oxylabs (Vilnius)	$43.7M ARR · 397 staff · $131M valuation, no VC	The mirror: same city, same bootstrap story, 40× the valuation
Decodo, IPRoyal, SOAX (Vilnius-heavy)	Mid-market, $2–7/GB	The neighborhood is the world capital of this industry
NetNut (Alarum)	Was a certified top-8 incumbent	Seized 07-02 — supply was the Popa botnet (~2M hijacked devices); ⅓ staff cut. Proof that certificates ≠ consent. Displaced demand = hypothesis with selection risk — the fleeing buyers may be the ones to refuse
Grass	~$304M cap · 2.5M nodes claimed	Owns "ethical AI-data" narrative at ~100× MN's scale — do not fight it head-on
Browserbase · Anchor · Firecrawl	$15–40M rounds, agent-infra natives	The funded layer above raw IPs; MN enters that knife-fight last
Proton	Profitable on subscriptions; foundation-owned	The living "profit funds mission" precedent (structural, not promised)

Where MN might win (hypotheses, not proofs): buyers the incumbents structurally cannot serve — crypto-native, protocol-level access — and provable consent as an operating standard once counsel-reviewed. Zero named prospects or paid pilots exist for any of it yet. Where entering is suicide: generic proxy sales, agent-browser infra, consumer VPN growth.

Waves — the 2026 horizon board

The origin lesson: MN caught two fast waves and missed the two slow ones that sustain a company. Forward filter: one fast + one slow-structural.

Wave	Stage	MN grip
Agents-with-wallets / machine payments	Rising, early	Yes — 4.5yr of machine-clients-paying-per-use (ride x402/L402, don't rebuild)
Proof-of-personhood / provenance	Rising hard	Partial — consented-household network as the "real origin" layer (build-heavy)
Digital sovereignty / splinternet	Rising, geopolitical	Mission-native; WC-2 grants monetize it partially (OTF, $10k–900k awards)
DePIN — bandwidth	Rising sector, lowest-value tier	Native — and the least money; don't lead with it
DePIN — compute/GPU	Revenue leader	None — closed, don't chase

Transport reality check: MN's own CEO reaches the western internet from Russia via HAPP (VLESS+Reality), not Mysterium — the censorship niche needs a transport build, and the layer that wins censorship is given away free. Feature and credibility play, never a business model.

Risks

Two registers: Lee's exposure, and MN's structural hazards.

LeeBrand contamination. Strategist to a residential-proxy network in a category with FBI-scale takedowns and German node-runner home raids. No public use of Lee's name without written approval — now a pre-signature gate (clause 3).

LeeŠaras single-channel. Money, insider data, and proposal ownership flow through one co-owner whose structural incentives are not identical to Lee's. Fix: the Lee–Šaras side agreement (clause 10), not suspicion.

LeeWork is running (since 08-10) with the agreement unsigned — Robertas himself notes it in the 08-11 September message. Payer named only as "Ro's MB attached to MN"; entity name/code, signer, payment mechanics and upside authority still open; clause set drafted, Lee-gated. Trust is real (largely via Šaras) — the paper still is not. And the upside clock runs one way: negotiating a share after showing MN the winning opportunity is negotiating face-up; today there are candidates and no winner — maximum leverage, closing (08-12).

LeeLee's own entity undecided. Individuali veikla vs thawing the dormant MB — at ~€60k/yr run-rate both the €42.5k credit taper and the €45k VAT line are crossed; the LT-debt withdrawal-timing question makes this an accountant call, not a default. One paid hour before invoicing.

LeeOpportunity cost. Content streak at zero since ~06-21; pre-proposal research had no hour cap. The compounding engine paid for this detour. Research is frozen: paid work only from 08-10.

Lee"AI slop won't be tolerated" is now a promise to a decision-maker. Any generic Kairos artifact is the exact failure Gitanas named. Show mechanisms, not assurances — a mechanism makes skimming expensive; a policy only forbids it.

MNIndifference, not nonpayment. 15-year-runway calm + money-nonchalant founder = deliverables can land in a void. The sprint forces written decisions (day 8) precisely because of this.

MNThe melting base — if real. If −10.2%/mo is secular, VPN revenue halves in ~7 months. But the rate is n=1; the 12-month series is the first paid pull.

MNGoProxies supply mismatch. 80M+ advertised IPs vs ~22k–31.5k network nodes. Not an allegation — an unreconciled gap that gates every proxy-scaling thesis and could hide upstream-supply risk.

MNNode-runner liability. Indemnity points at the operator (NetSys terms, Panama law); German raids documented; the earners are the exposed. Structural.

MNSingle-source financials. €30M / €2M burn are one verbal channel (Šaras, 07-10). Treasury may be crypto-denominated — the calm may be marked-to-market. The 08-07 deck's $288k/mo against the old $75.6k recollection shows how wide the perimeter fog runs.

Open questions — what gates everything

Canon lives in questions-to-excavate. Movement since 07-18 is marked inline — a partially answered question shows what closed and what remains, so nothing stands as "unknown" that no longer is.

Which entity signs? PARTLY CLOSED 08-01/08-03 — payer = Ro's MB "attached to MN"; Lee contracts into it under his own structure. Still open: the MB's name/code, what "attached" means legally, how it squares with the fund/mothership funding line, and which entity can grant upside.
Which entity invoices GoProxies customers? OPEN — no public source can close it; first extraction ask of the paid month.
Commercial vs mission perimeter OPEN — now urgent — the 08-07 all-hands made it visible: deck total $288k/mo vs the old ~$75.6k recollection is a perimeter question, not a growth story. Sprint deliverable #1; board-level.
VPN decline: fixable or secular? SHARPENED 08-07 — subscriptions recovered to May level; the −10.2%/mo melt downgraded to one earlier observation. Fork still open; the 12-month cohort/churn series stays the pivotal pull.
GoProxies supply map OPEN — per-SKU source, transfer price, consent evidence, quality, contribution margin. Answers the reseller fork. New wrinkle: SuperProxy's July "win" is 78% paid tests.
GoProxies concentration OPEN — 3 whales or 40 accounts? Different scaling plans entirely.
Real burn + breakdown, and whose money SHARPENED AGAIN 08-12 — the "burn ~€2M/yr" recording was corrected: Šaras's words were revenue ~€2M + loss ~€2M, so the cost base is ~€4M/yr derived, breakdown unknown. Kairos itself is paid from the fund/mothership, outside Ieva's ring. Still open: BlockDev burn ownership, treasury denomination/owner/deployment authority.
Core values, precise wording OPEN — the one hard bound on every pivot.
What exactly did Gitanas challenge? PARTLY RECOVERED 08-03 — doubt #1 recovered: "don't just report the problem" (answered by the torpedo method). The rest of his objection map is still the cheapest read on where board doubt lives.
Does MN have any central data layer? OPEN — sharpened 08-11 — a BigQuery warehouse exists (direct access 08-11: curated claude.* tables beside a duplicate-carrying looker/BI feed), so the question narrows: is it the org's shared ground truth, or one team's plumbing that the integrity defect shows nobody reconciles? The cheap falsifier stands: ask where an operator finds last month's numbers for a function they don't own. Gates the org-hub mammoth and explains the marketing standard gap.