Kairos · the weekly one-pager

Every Friday, the same page.

What we did, what we found, what we got wrong, and what we need decided. One page, every week, whether the week was good or not.

Latest · week 5 · What the company can and cannot measure 11 September 2026 Šarūnas Savickas · Linas Matulis
How to read this page

Six sections, the same six every week, in the same order. Read them in order and you have the week; read only What we need from you and you have the part that is on you, ranked by what it unblocks and dated. Every finding carries a status chip saying how much weight it can carry, and any underlined term opens its own plain-language definition. Nothing here needs you to have been in our sessions.

Every finding carries its status InsiderVerifiedReportedInferenceHypothesisKilled
Week 5

What the company can and cannot measure

5–11 September 2026 Call · Friday 11 September 2026, 13:00 Published 11 September 2026

Month one ends today. The sprint ran from 10 August to 11 September, and this edition is the written record the review reads from.

This week the work turned from Mysterium's data to Mysterium's own reporting. We read every all-hands deck from January 2024 to September 2026, opened the picture layer that earlier passes could not see, and separated two questions that had been running together: what the company stopped telling itself, and whether the numbers still exist to tell. Alongside that we read the customer estate through five independent openers, closed a seven-day investigation into where a customer's intention survives a transition, and built a working demonstration of a complete consumer product.

None of it is a commercial outcome. No intervention has run, and no revenue, retention, reliability or cost figure has moved. This edition reports evidence, corrections, and the decisions the evidence now makes available.

What we did

Artifacts produced, heaviest job first — not a list of activities.

1. Read the all-hands deck series end to end, including the picture layer.

  • Opened all 33 dated decks as OOXML. There is no machine-readable chart layer anywhere in the corpus: zero native chart parts and zero embedded workbooks. Every chart the company has shown its staff in three years is a flat picture.
  • Read all 187 unique images, by optical character recognition and again visually, each visual read done by a reader given a neutral extraction brief and no hypothesis. Until that pass ran, every claim about a missing metric was an upper bound rather than a fact. It is now closed, and closing it moved one answer.
  • Separated disclosure from capability for each metric: whether the deck series once showed it, and whether the warehouse can produce it today. Paid acquisition and the cost side answer one way; churn, retention and lifetime value answer the other. Paid acquisition is the harder of the two.
  • Named the three ways the finding could have been wrong and tested each: a cost metric hiding inside the 2025 and 2026 chart images, a separate finance forum carrying the reporting, and a corpus with missing decks. The first moved the date by which reporting stopped. The other two are refuted.

2. Read the customer estate through five independent openers, and produced the customer evidence the product argument rests on.

  • Coded every free-text answer in the in-app cancellation survey against one codebook, and dated the wrong-plan complaint to a specific product change.
  • Ran twenty blind stratified customer journeys across six keys, naming the screened and excluded cases rather than dropping them.
  • Completed the paying-account map: every qualified 2026 closed-cycle account dispositioned, with the blind support audit run to completion and its coverage recorded beside it.
  • Extracted the company's growth-metrics workbook twice, blind and independently, then red-teamed the difference on denominators before anything was reported from it.
  • Reconciled two warehouse sources that disagree about refunds, and located most of the money difference on dates that only one of the two sources returns.

3. Closed the seven-day investigation into where a customer's intention survives a transition, which is also where the product reliability evidence came from.

  • Tested what survives a replacement purchase, a subscription resume, an address refresh and a changed observation window. Those are the transitions where the customer has already decided and the product still has to deliver.
  • Reproduced, at pinned client source, a case where accepting the resume prompt drops the customer's requested exit address from the retry. A one-line candidate retains it. Opposing review accepted the source reading and did not establish installed-build behaviour, so it stays unadopted.
  • Established what the public node listing does and does not measure, after our own two-hour experiment produced a result the instrument could not support.
  • Built and browser-tested a private matched comparison of two retry designs, and ran offline playback experiments on public traces to test whether a warning can be given early enough to act on.

4. Rebuilt the residential-versus-datacenter job test after an opposing session found a power defect in it.

  • The plan reached its third version through three blind review rounds. A fourth reader then showed that at the only published base rate, the test as designed could not have detected the effect it was built to find. All three earlier rounds missed it, and so did we.
  • Inverted the selection rule: pick the task classes on documented operational blocking rather than on terms of service, after verifying at source that one major service's terms restrict geography and never mention proxies at all.
  • Dropped one whole task class on a design argument rather than a measurement one. Repeated requests from rotating residential exits against a third party's booking systems is a cost we would be imposing on someone who never agreed to the experiment.

5. Took the network's direct product economics apart from first principles.

  • Derived the physical floors of a residential-address VPN, wrote the cheap falsifier under each, and checked them against our own source and registry findings rather than against marketing.
  • Compared what the network pays its operators against the capacity those operators have installed. Paid usage sits orders of magnitude below that capacity, and the gap held at the most generous corner of a sensitivity grid. The capacity side counts earning identities, which are not the same thing as independent households, so it is an upper bound.
  • Ran it past two peer sessions and landed both sets of corrections, including one that walked back two of our own overclaims about routing and one that disqualified vendor valuation figures as evidence that residential access works. The same rule disqualifies Mysterium's own marketing.

6. Built a working demonstration of a complete consumer product.

  • Elsewhere shows one continuous journey: arrival, purchase, use, a recoverable interruption handled without asking the customer to retry, and the end of the job. It runs in a browser and needs no sign-in.
  • It is a demonstration, not a proposal for a venture and not a commitment anyone has made. Its connection outcomes are authored; no network performance is measured by it, and nothing in it is an MN production change.
  • The point it is built to make is narrow: the work a customer should never have to notice is the product, not a missing button.

7. Kept the data estate and the delivery surfaces working, and put a comment layer under everything we ship.

  • The one-page opportunity portfolio for the September board meeting is live behind access and reads correctly for the core tier.
  • The complaints work now ships as a ranked briefing rather than a certified study, after the first version was read and found unclear.
  • Root-caused the daily brief's staleness to a disk exhaustion that silenced the scheduled run, repaired the failure path, and red-tested the repair.
  • Diagnosed three co-timed monitoring alerts as one cause, and not the cause they named: the machine had lost its uplink. The watches now separate an offline machine from a dead instrument.
  • Built Margin, a mark-and-comment layer, across the War Room. A reader can now mark a passage and answer it where it stands, so access and delivery of a finding no longer end at reading it.

What we found

Each line opens with how much weight it can carry.

  • Verified The cost side left the company's own reporting in January 2025, and the picture layer is where the last of it was hiding. Acquisition cost carried a numeric target with monthly actuals through 2024; a single January 2025 slide set revenue against costs; nothing after it does. What survives of paid acquisition reporting is three adjectives in twenty months, and the most recent one glosses the term for its own audience.
  • Verified Acquisition cost is not producible from the warehouse today. No cost, spend, budget or ad-platform column exists in any dataset, no pipeline has ever landed spend, and the attribution table that holds the other half of the ratio now assigns almost every purchase to direct or none. This is a capability finding, not an accusation, and it is the stronger of the two because it does not depend on anyone's motive.
  • Verified Churn, retention and lifetime value are the opposite case. Those tables are live and were refreshed the same week the deck was presented, and a monthly consumer-VPN churn series is one join away from two of them. Not showing them is a choice; not showing acquisition cost is a limit.
  • Verified A reported year-on-year improvement in the staff survey is a response-rate artefact. The score rose while the number of favourable respondents fell, because a large part of the previous year's respondents did not answer at all. Participation was disclosed on its own slide; the disclosure and the comparison were kept apart, and a shrinking response base is not a comparable series.
  • Verified A flow presented as releasing next week was not in production three days later, on the company's own task board. The survey that asks people why they cancelled has been collecting answers for about fifteen months while the flow to let them cancel was still being built.
  • Inference The claim that acquisition is now bringing higher-quality users is equally explained by a shift toward longer prepaid terms, which books more revenue per transaction without any change in the customer. It is decidable against the warehouse by decomposing year-on-year August revenue per transaction by plan term. That decomposition is pre-registered and its method passed a blind cross-family review this week; it runs next. If term mix and price carry the lift, the claim currently justifying the acquisition strategy is unevidenced.
  • Inference What the network pays its operators, set against the capacity those operators have installed, implies utilisation far below one percent. If it holds, the node network is not a supply constraint waiting to be relieved but an asset with almost no demand against it. Two things would kill it. Off-chain revenue invisible to an on-chain decode would have to run more than a hundred times the decoded settlement before capacity binds at all. And the capacity side counts earning identities rather than independent households, so a mapping that collapses many identities onto one connection closes the gap directly. Opposing review accepted the observation this week and refused the stronger reading, so demand limitation stays a hypothesis beside geographic scarcity rather than a conclusion.
  • Verified The public node listing can omit an eligible node simply because a different subset was returned. The endpoint applies a per-country record cap while iterating an unordered map, and filters afterwards. A direct lookup returned every one of a seeded sample of nodes the listing had just omitted.
  • Verified The exhaustion notice a customer sees when refreshing an address does not measure exhausted supply. The client emits it when successful refresh registrations reach a location count, without inspecting distinct returned addresses or requiring a failed attempt.
  • Killed Mysterium's residential inventory as an uncontested network-measurement business does not survive its incumbents. One vendor already sells measurement from end users' home networks, and another excludes VPN-hosted probes by design because tunnelling distorts the route being measured. What remains is narrower: selected additional vantage points for an existing buyer, conditional on node-local execution, origin and measurement quality being demonstrated.

What we changed our mind about

What we got wrong, printed next to what replaced it. The week this section is empty is the week to distrust the page.

  • Our own two-hour node-availability experiment was measuring the wrong thing. The apparent disappearance of nodes between samples was the listing endpoint's sampling behaviour, not provider churn, and the geography of the effect is what exposed it. The reliability reading is withdrawn; the falsification and the retrieval rule that follows from it are the results that stand.
  • The residential-versus-datacenter test plan had a power defect that three blind review rounds missed, and so did we. A test that cannot detect the effect it is built to find returns a null that says nothing about the world, so the null we would have reported next week would have been about the instrument, not about residential addresses.
  • Our reading of the 2024 advertising-return figures took a threshold row for a monthly actual. We re-read the raw slide markup for the three months concerned and corrected it. The finding it sat inside was unaffected; the number was not ours to get wrong.
  • A cancellation-cause table was sorted by row count while displaying user count, which inverted the top pair. Something breaking leads, not the subscription no longer being needed. The sort is corrected, and so is the prose that rested on it.
  • A peer session's conclusion that supply cost consumes most of revenue used a revenue baseline our own facts board strikes through. At the measured denominators it is a far smaller share, and we reconciled the rate contradiction rather than letting the two numbers stand side by side.
  • Last week's paid-acquisition ask is withdrawn, because we answered it ourselves from a source we already held. A current internal workbook shows material weekly paid spend by product, which settles the fork the ask existed to settle. What it does not carry is the channel and campaign split, and we will ask for that only when a specific allocation decision needs it.
  • Last week's attribution-owner ask is parked for the same reason rather than reprinted. We have reached the ceiling of what internal evidence can establish about the attribution lineage; the remaining facts are live platform configuration, which is worth a request only against a decision.
  • The Pipedrive ask is reprinted below rather than quietly dropped. It has gone unanswered for two weeks, and the credential still does not authenticate.

Next week

Committed, and checkable against next Friday's edition.

  • Run the plan-term decomposition on year-on-year August revenue per transaction and report whether the higher-quality-users claim survives it, including if it does not.
  • Run the rebuilt residential-versus-datacenter test at its corrected base rate, one variable at a time, and report the power calculation beside the result.
  • Close the seven-day investigation's remaining transition questions with a changed contract or a genuinely different source, not another catalogue entry or another synthetic click-through.
  • Take the customer estate's five reads into one connected customer picture, preserving the distinctions between contacts, accounts, people and outcomes.
  • Install the comment layer on the remaining Kairos pages, so that every finding we ship can be answered where it is stated.

What we need from you

Ranked by what it unblocks. Named person, exact thing, by-when.

Who owes whatRobertas14Šarūnas235

first one with a date of its own · Šarūnas, Wed 16 Sep — ask 2 (then 4 more). The order below is what each one unblocks, not when it is due.

  1. Robertasby the end of today's review

    the continuation decision, its scope, and who owns it. Month one ends today, continuation was always conditional on it, and everything below assumes an answer to this one.

  2. Šarūnasby Wed 16 Sep

    the VPN and SuperProxy revenue split for August. The 4 September revenue bridge does not close from the deck's own figures, and the split is the one thing that would close it.

  3. Šarūnasby Wed 16 Sep

    the replacement Ahrefs key requested on 5 September. The existing key has not authenticated since 27 August, and the search room reports a dead source every day it stays that way.

  4. Robertasby Wed 16 Sep

    confirmation of whether the cancellation flow is in production, and the name of the person who can confirm it. The board saw it as releasing; the task board three days later did not.

  5. Šarūnasby Wed 16 Sepcarried from week 3

    the read-only Pipedrive grant. The current credential still returns an authentication failure, and the CRM room stays uncommissioned until it lands.

Notes

The caveats that belong on the record, not in the call.

Every figure behind these findings — the deck values, the survey counts, the warehouse denominators, the customer records and the reviewer receipts — stays in the protected Kairos room and travels in the call. This public weekly carries the decisions, the evidence limits, the corrections and the asks.

The commitments under Next week are ours to keep and are checkable against next Friday's edition. They assume the engagement continues past today's review; if it does not, this edition is the closing record rather than a midpoint.

The mechanical sweep first counted every dated record inside the seven-day window, then selected records carrying both a Kairos or Mysterium scope term and a landed-work term. Classification is first-match-wins in the published rule order, and every selected record carries its matched scope, landed and rule terms without raw log text. Two new included groups were added this week for the seven-day investigation and the all-hands deck series; three new parked groups cover machine upkeep, private working notes and agent plumbing, none of which produced a client artifact.

Week 4The map started producing decisions29 August–4 September 2026
Week 4

The map started producing decisions

29 August–4 September 2026 Call · Friday 4 September 2026, 13:00 Published 4 September 2026

Week 4 turned the evidence map into a sharper decision system. Kairos isolated one paid-acquisition destination with no recorded web purchase in the primary tracker, corrected the direct product economics, replaced install-based customer stories with account-level tests, reproduced bounded VPN protection failures, and re-ranked the opportunity portfolio. The protected War Room carries the detailed evidence and denominators behind those decisions.

The business has not yet acted on one of those decisions in a way that produced a measured commercial or customer outcome. This edition therefore reports the decision changes, the evidence limits, and the next tests. It does not book revenue, retention, reliability, or cost improvement.

What we did

Artifacts produced, heaviest job first — not a list of activities.

The shipped work below is ordered editorially. The seven sweep categories map into these reader-facing outputs; the complete denominator and selector receipt remain adjacent to this source file.

1. Completed the paid acquisition investigation.

  • Traced the Take Back destination through the primary server-side tracker, a separate GA4 lineage, and the available campaign-name fields.
  • Found no recorded web purchase in the primary tracker or the GA4 slice. Kept the conclusion below a return-on-investment verdict because campaign spend, incrementality, app-store outcomes, and most campaign-name coverage remain unavailable.

2. Rebuilt the customer evidence and retention questions at account grain.

  • Replaced install counts as a proxy for people with account-level membership rules and prospective outcome tests. Heavy accounts remain compatible with households, teams, resellers, device churn, or abuse.
  • Replayed the customer journey across acquisition, product use, support, retention, economics, and learning. The replay made the missing joins visible; it did not manufacture a complete customer journey from partial records.
  • Kept Payment Recovery inside Retention because an eligible continuation-intent cohort and incremental retained contribution remain unmeasured.

3. Turned product reliability observations into runnable falsifiers.

  • Reproduced cases where the Mysterium VPN client state and the usable protected path disagreed on one controlled Mac.
  • Converted those observations into direct, datacenter, residential, failure, recovery, leak, and teardown tests. One environment establishes a mechanism, not customer incidence or churn causality.
  • Ran a separate free-plan competitor benchmark to improve the test design. It remains a constrained comparison, not like-for-like proof about Mysterium.

4. Re-ranked the opportunity portfolio and corrected direct product economics.

  • Corrected the mission boundary before ranking. Paid-acquisition reallocation is first, heavy-account and device-slot economics are second, and the technical-failure and protection-truth cell third. Slots four and five stay empty because no other candidate clears the current evidence bar. None is an approved bet.
  • Removed direct-proxy conversion and cross-sell from this VPN-plus-adjacent mission rather than pretending they were killed. Take Back as currently configured was rejected as a growth thesis; reallocation of its resources is the surviving measurement question.
  • Framed Real Access as a provisional VPN learning commitment: prove that a truthfully labelled residential exit completes ordinary lawful tasks where a datacenter exit fails. Product, pricing, launch, and pursuit remain outside this report's authority.
  • Corrected the direct product economics so gateway fees, traffic, and refunds no longer blur into profit. The selected direct lines leave contribution; payroll, advertising, tools, infrastructure, and group costs remain outside that measure.

5. Advanced the complaints study without promoting unfinished findings.

  • Completed the classified corpus and blind-coder comparison under the certified research architecture, then kept the findings in adversarial repair when provenance and adjudication checks failed.
  • The current round has not produced a terminal publishable verdict. No complaint frequency, cause, segment, or commercial conclusion is promoted here.

6. Strengthened the data estate, access and delivery controls.

  • Repaired unattended MN Daily capture and verified the installed service path, not only an interactive run.
  • Kept private rooms behind their existing access tiers while making aggregate decision surfaces reachable to their intended readers.
  • Continued to separate source access, write protection, coverage, decision fitness, and delivery. A green login or a deployed file does not collapse those states into one claim.

What we found

Each line opens with how much weight it can carry.

What we changed our mind about

What we got wrong, printed next to what replaced it. The week this section is empty is the week to distrust the page.

  • The campaign conclusion no longer rests on a broad “four instruments” claim. The primary tracker and one GA4 lineage support the bounded zero; sparse campaign-name fields now define an attribution limit rather than a confirming instrument.
  • A large install count no longer counts as evidence of many customers or a valuable segment. Account membership and outcomes must be tested prospectively.
  • “Connected” no longer counts as the customer outcome. Protection requires the tunnel, public route, DNS, and usable data path to agree.
  • Direct contribution no longer counts as profit. The missing operating costs stay visible instead of being implied away.
  • A completed analysis no longer counts as business impact. No outcome is booked until an authorized intervention changes a measured result.
  • A large classified complaints corpus no longer counts as a publishable study. The evidence package must survive provenance, adjudication, and terminal review.
  • Last week's ClickUp safety ask remains open. The current doorway passed a live read, but its personal token remains source-write-capable. Kairos-side code that permits reads and refuses writes proves only our adapter boundary, not vendor-enforced write denial. The read-only identity or MN-controlled broker, backup owner, and replacement and revocation rehearsal are reprinted below as one carried ask.
  • Last week's Pipedrive and affiliate-attribution owner asks also remain unresolved; both are reprinted below as carried asks rather than silently reset. The current Pipedrive credential still returns an authentication failure, and the named attribution-lineage owner remains unrecorded.
  • Last week's Robertas ask had two parts. The opportunity decision-room part is withdrawn because the corrected mission has three provisional ranks and two open slots; there is no decision-ready candidate for an Opportunity Schedule. The month-one review and circulation date remains open and is carried below.

Next week

Committed, and checkable against next Friday's edition.

  • Obtain campaign cost by campaign and month, bind it to the primary tracker, and decide whether a bounded reallocation test is warranted without moving spend first.
  • Close the complaints study's provenance and adjudication findings, obtain a terminal opposing verdict, and promote no finding before it clears.
  • Run the controlled direct, datacenter, residential, failure, recovery, leak, and teardown sequence with one variable changed at a time.
  • Test the customer membership rules prospectively and preserve the no-action comparator; do not promote a customer segment from the current cells.
  • Keep opportunity slots four and five empty until a contender proves its own mechanism, reachable population, magnitude, and contribution path. Do not open an Opportunity Schedule from vacancy or from the current E1 signals.

What we need from you

Ranked by what it unblocks. Named person, exact thing, by-when.

Who owes whatŠarūnas1234Robertas5

first one with a date of its own · Šarūnas, Tue 8 Sep — ask 1 (then 4 more). The order below is what each one unblocks, not when it is due.

  1. Šarūnasby Tue 8 Sep

    route the Google Ads campaign-cost export by campaign and month, plus the owner of the campaign-to-purchase mapping.

  2. Šarūnasby Tue 8 Sepcarried from week 3

    route a vendor-enforced read-only ClickUp identity or an MN-controlled read-only broker; name the backup owner and rehearse replacement and revocation.

  3. Šarūnasby Tue 8 Sepcarried from week 3

    route the current read-only Pipedrive path or the account owner who can grant it, so the existing CRM room can be commissioned.

  4. Šarūnasby Tue 8 Sepcarried from week 3

    return the named owner and mapping contract behind the warehouse marketing-attribution table, including how affiliate orders should reach its classification.

  5. Robertasby Fri 11 Sepcarried from week 3

    confirm the month-one review and circulation date.

Notes

The caveats that belong on the record, not in the call.

Detailed figures, queries, source inventories, customer records, and reviewer receipts remain in the protected Kairos room. This public weekly carries only the aggregate decisions, evidence limits, corrections, commitments, and asks. The mechanical sweep first counted every dated record inside the seven-day window, then selected records containing both a Kairos/MN scope term and a landed-work term. Classification is first-match-wins in the published rule order; each selected record carries its matched scope, landed, and rule terms without raw log text.

Week 3Data access became an operating system22–28 August 2026
Week 3

Data access became an operating system

22–28 August 2026 Call · Friday 28 August 2026, 13:00 Published 28 August 2026

Week 3 moved data access from a collection of credentials into one visible operating system. Every source now passes through the same path: register the room, prove a useful read, prove write protection, measure the reachable perimeter, commission one named use, and keep the connection current. The Data Access Gold Standard defines the contract. The Data Room map shows the evidence, the current gate and the exact next action for every room. The known ClickUp workspace perimeter is now readable, Impact is producing a useful read after an exact query correction, and Pipedrive has a named access blocker. The map keeps useful data, write protection, coverage and decision fitness separate instead of flattening them into one access label.

That access work has started producing decisions rather than inventories. We completed a campaign operating review, built the first customer-intelligence layer from a bounded support corpus, and opened an opportunity evidence programme that compares every candidate against the available source rooms before changing the portfolio order. The current top five remain provisional until every candidate has been checked against every available source room.

Alongside it, we built the Themis legal architecture and the Jubal case practice over the weekend, then published the internal judgment protocol presentation. Kairos Case 1 applies them to the lawful-data perimeter without exposing MN records or promoting scoping research into legal clearance. Key Maker and Delight remain evidence candidates inside the Mammoth Protocol gate. A signed Opportunity Schedule is required before material pursuit.

What we did

Artifacts produced, heaviest job first — not a list of activities.

Nine output groups moved decisions this week, in order of weight. The order is editorial; the denominator comes from the mechanical sweep.

1. Built the Data Access operating system.

  • Published one acceptance standard for every source connection. Authentication, useful data, write protection, coverage, decision fitness and durability now carry separate evidence states.
  • Published the Data Room map and its machine-readable record, then added a room-by-room onboarding board. A readable source can no longer appear blocked because its safety boundary is unfinished, and a received credential can no longer appear ready before the real work succeeds.
  • Commissioned the complete known ClickUp perimeter, repaired the Impact query, classified every reachable Omnisend campaign, converted Ahrefs evidence into a bounded verification queue, and registered Pipedrive with its exact blocker. None of those source reads bypasses the separate write-protection gate.

2. Opened the opportunity evidence programme.

  • Replaced the provisional opportunity order with one complete matrix of every candidate against every available source room, with explicit evidence gaps and prove-or-kill contracts. The current top five do not become final merely because they were first on the board.
  • Published Opportunities v0.8.1 with Key Maker visible as the first contender to replace the fifth slot. The leaderboard did not change, and the page keeps the promotion, park and kill conditions visible.
  • Split Key Maker into an MN operating enabler and a possible external product. Both remain at claim-assessment stage; neither has a proceed verdict or build authorization.

3. Built the customer-intelligence layer and campaign operating review.

  • Built a privacy-bounded support-conversation corpus and deterministic analysis that can be reproduced without publishing raw conversations or person-level data.
  • Identified recurring VPN need states and partial journey fragments, then named the missing customer, product, outcome and economic joins. Support evidence does not establish demographics, power users, retention or revenue.
  • Classified the complete reachable Omnisend campaign estate into keep, finish, retire and investigate decisions. Performance winners remain unknown because the current source does not prove outcome, audience or conversion effects.

4. Completed the affiliate-attribution integrity control.

  • Compared independently proven CJ affiliate orders against the warehouse purchase and marketing surfaces across a correction-safe monthly window.
  • Closed the comparison with an attribution conflict: the orders reach the warehouse, while affiliate classification is absent. The existing Slack corpus did not identify a source-addressable lineage owner, so the repair now requires the named source owner and mapping contract.
  • The same comparison now runs as a repeatable monthly control.

5. Opened the MN Slack evidence layer.

  • Built a local full-text reader over the supplied archive, then published an owner-only War Room room after removing credential-shaped spans. The raw archive and generated search data remain off Git.
  • Produced the first full channel map, MN operating-team map and payment-recovery instrument from the corpus. The instrument separated payment-recovery mechanics from customer intent and corrected a misleading external benchmark denominator.

6. Built Themis/Jubal, then put the lawful-data perimeter through it.

  • Built the legal competency architecture through 52 recorded adversarial rounds over Saturday and Sunday. The accepted core fixes the authority hierarchy, evidence gates, hostile review, calibration and counsel boundary; its measurement appendix remains frozen under a separate unfinished review loop.
  • Defined Themis as the stable constitutional architecture and Jubal as the case practice that applies it. Neither is legal authority or a substitute for licensed adjudication.
  • Opened Kairos Case 1 around one useful question: which MN data may be accessed, by whom, for what purpose, through which path and under which controls.
  • Built a 22-source pinned shelf and the v0.2 scoping memo. Its citation gate verified all 66 quoted blocks. The memo uses abstract data classes only and withholds its restrictive and permissive readings—and any lawful-access conclusion—until the minimum facts land.
  • Built and published the internal Themis/Jubal judgment protocol presentation behind the existing Kairos access boundary. Recipient use remains unmeasured.

7. Defined Key Maker without promoting it.

  • Published the internal Working Backwards release, technical architecture and interactive switchboard.
  • Defined the future acceptance conditions around complete operating context, visible unknowns, role-bound authority, verified onboarding and effective access denial. The package states its fictional future status and its Mammoth boundary on the page.
  • Reopened the candidate from first principles after separating the MN-enabler and external-product propositions. Passive evidence can accumulate through normal source work, but one coordinator journey is only a possible evidence unit—not a preselected next action.

8. Opened the VPN evidence room and Delight.

  • Published the VPN testing room with the current first-hand observations, evidence limits and next sizing sequence.
  • Published Delight as a working opportunity inside Retention Economics. Reconnect, refresh-IP and node-replacement machinery already exists; the opportunity is to close, harden and measure the recovery loop. Production prevalence, retained value and a Mysterium-specific advantage remain open.

9. Kept the evidence boundary and operating machinery working.

  • Repaired the daily-brief scheduler on Saturday, rebuilt the missed editions and renewed the daily evidence surface through the end of the reporting window.
  • Hardened the Kairos repository and autonomous agent rail against detached work, stale claims, failed report publication and ambiguous message state. These are delivery controls, not new opportunity claims.
  • Moved owner-only source material behind identity-specific room policy and tested the room boundary against direct paths, encoded aliases and shared search surfaces.
  • Kept client-facing pages focused on aggregate findings, source state, decisions and asks. Raw source material and person-level judgments stay out.

What we found

Each line opens with how much weight it can carry.

  • Verified A source can return useful current data while its write protection, coverage, ownership or decision fitness remains unfinished. Those states now stay separate on the operating map.
  • Verified CJ-proven affiliate orders reach the warehouse without affiliate classification. The current acquisition view therefore cannot support an affiliate-performance decision.
  • Verified The commissioned Slack history does not name a source-addressable owner for that mapping. Repeating the same corpus search will not repair the lineage.
  • Verified The complete known ClickUp workspace perimeter is readable through the authorized identity. The remaining blocker is safety: the current token is not vendor-enforced read-only, so the source is not fully commissioned.
  • Verified Impact’s empty result was caused by a missing report selector, not a missing data entitlement. The corrected query returns useful current data; coverage, write protection and repeatability still keep the source short of full commissioning.
  • Verified Pipedrive remains source-blocked. The current local credential does not authenticate, and the next action is an exact read-only grant from the account administrator—not another broad access request.
  • Verified Grafana and Ahrefs return current source data. Their current access paths still lack an accepted server-enforced read-only boundary, so no business finding from either source has been promoted.
  • Verified The reachable Omnisend campaign estate can be classified into operating actions. It cannot yet establish campaign performance or causal impact because analytics, message-body and stable campaign-to-work joins are incomplete.
  • Verified The support corpus shows recurring need states and partial VPN journey fragments. It does not identify demographics, power users, a complete journey, retention effects or revenue impact.
  • Verified The current opportunity order is provisional. A complete candidate-by-source audit must close before the first evidence wave can confirm or replace it.
  • Verified The lawful-data perimeter now has a current source shelf and scoped matrix, but no lawful-access conclusion. The in-house blind framing and the minimum facts must land before the matrix proceeds; licensed adjudication remains a separate gate.
  • Inference Outcome-verified VPN self-healing belongs inside Retention Economics. Production incidence, support burden and retained-value impact decide whether it deserves a separate opportunity slot.
  • Hypothesis Key Maker may become portable beyond MN if repeated access work shows that governed credential operation is a separate, recurring buyer problem. The current evidence does not yet prove that market.

What we changed our mind about

What we got wrong, printed next to what replaced it. The week this section is empty is the week to distrust the page.

  • A credential receipt no longer counts as access. A room reaches usable state only after its read, safety, coverage and first decision use have passed independently.
  • The Impact block was not a permission gap. The source had useful data behind an exact selector that the original query omitted; the administrator request built on the entitlement conclusion was withdrawn.
  • The marketing-attribution task begins with repair. A current instrument already exists, and its useful coverage is worsening.
  • Internal research and agent review do not equal legal clearance. Themis/Jubal now keeps the source shelf, both legal readings and adjudication state visible.
  • Delight is not a greenfield feature. It stays inside Retention Economics as the completion and measurement of recovery machinery that already exists. A separate opportunity requires evidence that failure exposure is material and changes support or retained value.
  • Key Maker does not get a preselected operational proof. The portfolio audit comes first; passive evidence continues through normal source work, and a standalone build remains unauthorized.
  • A source inventory is not a decision order. The opportunity programme now prioritizes evidence by its power to confirm, kill or reorder a candidate.

Next week

Committed, and checkable against next Friday's edition.

  • Close the complete candidate-by-source audit, reconcile the provisional top five and publish the resulting evidence contracts before starting a test.
  • Begin the smallest top-three evidence contracts only after that denominator closes, while keeping outreach, product work, spend and production mutation behind the signed Opportunity Schedule gate.
  • Put the attribution conflict in front of the source owner, recover the mapping contract and bind the monthly control to an accountable repair path.
  • Resume Kairos Case 1 when the in-house blind framing and minimum facts land; record that position as counterparty input, not adjudication, and keep raw MN records outside the legal research corpus.
  • Recover the referenced customer-journey artifacts and commission the customer, product, outcome and economic joins needed to test the support findings.
  • Keep Key Maker on passive evidence capture until the portfolio comparison identifies a discriminating test worth authorizing.

What we need from you

Ranked by what it unblocks. Named person, exact thing, by-when.

Who owes whatŠarūnas123Robertas4

1 answered in the call itself · first one with a date of its own · Šarūnas, Tue 1 Sep — ask 1 (then 2 more). The order below is what each one unblocks, not when it is due.

  1. Šarūnasby Tue 1 Sep

    route a vendor-enforced read-only ClickUp identity or broker, with a named backup owner for maintenance.

  2. Šarūnasby Tue 1 Sep

    route the exact read-only Pipedrive grant to the Account Admin; the current local credential does not authenticate.

  3. Šarūnasby Wed 2 Sep

    return the named owner and mapping contract behind the warehouse marketing-attribution table, including how affiliate orders should reach its classification.

  4. Robertasby Friday's call

    confirm the month-one review and circulation date, and whether the opportunity portfolio should land in the same decision room.

Notes

The caveats that belong on the record, not in the call.

Detailed source inventories, figures, queries and access evidence stay in the protected Kairos room. This public page carries the shipped outputs, decision state, evidence limits and next actions.

Week 2Audit closed; customer evidence and team systems live17–23 August 2026
Week 2

Audit closed; customer evidence and team systems live

17–23 August 2026 Call · Friday 21 August 2026, 13:00 Published 21 August 2026

Week 2 moved Kairos from broad discovery into decision work. We finished the warehouse audit, published its evidence, and set the current opportunity order: Payment Recovery first, Retention Economics second and GoProxies Conversion System third.

We also added evidence the warehouse cannot provide. We synthesized public customer reviews, started using Mysterium VPN ourselves, kept the shared Kairos repository online from Lee's MacBook during the Forgejo outage, and put the autonomous agent relay and new team research surfaces into operation.

What we did

Artifacts produced, heaviest job first — not a list of activities.

Seven outputs shipped: evidence and opportunity work first, delivery infrastructure after.

1. Finished the warehouse audit.

  • Accounted for every live warehouse object, tied the result to reproducible query evidence, and published the protected audit brief.
  • Separated mechanically checked data from owner-confirmed business meaning. Broad mapping is finished; definition work now applies only to metrics that can change a current decision.

2. Built the customer-reviews evidence layer.

  • Captured 410 dated written reviews from the Apple App Store, Google Play and two Trustpilot profiles as of 20 August. The evidence carries its public-source limits.
  • Published the searchable Customer Reviews page and team synthesis, then added the review findings as guardrails on Payment Recovery and Retention Economics.

3. Put the autonomous agent relay into unattended operation.

  • Loaded the two-way Claude–Codex relay on 17 August. It completed agent handoffs and review hops without Lee carrying messages between systems, then red-teamed itself through five rounds before unattended operation.
  • Published the ranked relay use-case portfolio and connected it to the wider Epistemic Federation work: how sovereign personal agents can collaborate on shared work without merging their private context.

4. Ranked and narrowed the commercial opportunities.

  • Published Opportunities v0.7.1: Payment Recovery #1, Retention Economics #2 and GoProxies Conversion System #3. The first two have the strongest measurable decision paths; GoProxies has a clearer operating diagnosis and incomplete commercial economics.
  • Cleaned the payment-failure denominator, tested the other gateways, split renewal by plan tier and country group, and mapped the freshness and fitness of every source used by the leading claims.

5. Started testing Mysterium VPN as users.

  • Opened a 30-day product immersion and completed the first activation, connection, performance and protection baseline.
  • Tested the native crash and fail-safe recovery path, datacenter and residential connections, plus public-IP, WebRTC, DNS and third-party classification checks. These observations remain provisional until repeated.

6. Launched a temporary Kairos Git repository on Lee's MacBook.

  • On 17 August, brought the shared repository back online from Lee's machine while Robertas's Forgejo host remained unavailable. Clone and push were proven end to end, Lee's working copy was synchronized, and the return path to Forgejo was preserved.

7. Expanded the War Room into a team operating surface.

What we found

Each line opens with how much weight it can carry.

  • Verified Broad warehouse discovery can stop. The remaining data work is a small owner-definition pass over the sources behind current decisions.
  • Killed Payment Recovery cannot yet cover the full payment base. The app stores retain their retry outcomes, and the other warehouse gateways do not expose a usable charge-level result.
  • Verified Refunds, disputes, first-time cards and failed renewals now carry separate populations and denominators in the leading payment analysis.
  • Verified Among 410 captured written reviews, connection reliability, cancellation and billing, and customer support are the most frequently tagged negative themes. Tags overlap, and the public review history is incomplete.
  • Verified Billing and cancellation also clear the preregistered threshold in a full year of support data, giving the retention work a transaction-side and customer-side measure.
  • Verified Higher-priced plan tiers do not renew worse in the readable cells. Country group moves more than price tier in the current cut.
  • Verified The main revenue table still labels the new web gateway as Stripe. Provider-level comparisons stay blocked until its authoritative source and label lineage are confirmed.
  • Inference The public-review synthesis and support evidence make customer experience a load-bearing test for both leading opportunities. The finding changes their validation plan; it does not change their current order.
  • Verified One first-hand session reproduced a native crash followed by fail-safe recovery and successful residential connection. The 30-day run will show whether those observations are isolated or recurrent.

What we changed our mind about

What we got wrong, printed next to what replaced it. The week this section is empty is the week to distrust the page.

  • A full warehouse audit did not justify a warehouse-wide governance programme. The better rule is smaller: start from a named decision and govern only the metrics that could change it.
  • The non-Stripe gateways will not enlarge Payment Recovery quickly. That path stays closed until MN supplies processor or store outcomes.
  • The tier churn gap does not currently support a simple price-pressure story. The next test holds plan duration constant and examines product conditions by country group.
  • GoProxies is larger than a paid-test cohort question. Its operating diagnosis is now clearer; missing revenue, margin, retention and delivery-cost evidence keeps it third.

Next week

Committed, and checkable against next Friday's edition.

  • Recover MN's existing definitions and trusted reports for the few sources behind Payment Recovery and Retention Economics.
  • Build one comparable loss matrix across the two leaders: stable denominator, retained value, natural recovery, controllable loss and intervention owner.
  • Complete the country-group survival test while holding plan duration constant, and continue the 30-day VPN product immersion.
  • Complete the new web-gateway reconciliation and price the GoProxies conversion problem.

What we need from you

Ranked by what it unblocks. Named person, exact thing, by-when.

Who owes whatŠarūnas14The payments owner2Gitanas3

first one with a date of its own · Šarūnas, Mon 24 Aug — ask 1 (then 3 more). The order below is what each one unblocks, not when it is due.

  1. Šarūnasby Mon 24 Aug

    a thirty-minute owner-map session for the payment, purchase, subscription, churn and attribution sources, plus introductions to the two candidate owners.

  2. The payments ownerby Wed 26 Aug

    retry rules, the meaning of the do-not-retry class, the authoritative source for the new web gateway, and one current processor dispute count.

  3. Gitanasby Mon 24 Aug

    monthly marketing spend by channel and one working source for the current marketing pipeline.

  4. Šarūnasby Wed 26 Aug

    the GoProxies paid-test and retained-conversion ledger, with revenue, margin and delivery cost; or confirmation that no usable record exists.

Notes

The caveats that belong on the record, not in the call.

Detailed audit and customer evidence stays in the protected Kairos room. This public page carries the shipped outputs, decision state, evidence limits and next actions.

Week 1The numbers had to be rebuilt first10–16 August 2026
Week 1

The numbers had to be rebuilt first

10–16 August 2026 Call · Friday 14 August 2026, 13:00 Published 14 August 2026

Governed read-only access to Mysterium's warehouse opened on day two, and the first thing it produced was not an insight but a correction: the revenue numbers we were about to build on did not survive their own check. We rebuilt them, and the rebuilt versions hold across three independently constructed sources. That is the week. A unit that had gone straight to conclusions would have shipped a confident page on Friday and been wrong by a wide margin.

Everything else was groundwork for the one deliverable this month points at: the ranked opportunity portfolio in front of your board in the first week of September — screened wide, validated narrow, launched only through an explicit readiness gate. The numbers came first because every candidate in that portfolio will be sized off them.

The method that will rank that portfolio also exists now, and it is the week's other durable output. Every load-bearing claim carries where its evidence came from and, separately, how good that evidence is. Kept apart, the two make a well-argued guess impossible to mistake for a measured fact.

What we did

Artifacts produced, heaviest job first — not a list of activities.

Three jobs this week, in order of weight.

Rebuilt the numbers the month stands on.

  • Verified governed read-only warehouse access end to end and ran the first reconciliation pass against the July figure reported in the 7 August All Hands.
  • Traced the problem to duplicate rows sitting in the reporting-facing revenue tables, rebuilt the twenty-four-month series off the clean curated table instead, and confirmed the rebuilt figures against two independently constructed sources.
  • Extended the reconciliation back through 2023 and 2024 using the transcript archive, so the deck-versus-warehouse question now has a three-year context instead of a one-month one.
  • Published a warehouse blindspot map: what this data can answer, what it cannot, and which question each gap is blocking, ranked by how much it matters.
  • Received and verified the recorded half of the All Hands archive: all sixty-seven meeting transcripts accounted for and checked. The slide decks are not in our hands yet, and no recordings exist for 2019–2020 or 2025–2026, so for those years the decks are the only record — the ask is below.
  • Corrected our own record of the cost base after re-reading the primary source, and propagated the correction through every note that had built on the wrong figure. The error was ours; the discipline that caught it is the one this unit runs on.

Laid the groundwork for the September portfolio.

  • Built the evidence ranking every candidate will be judged on, and put the whole board through it. A claim carries where its evidence came from — desk reasoning, company-reported, measured, reconciled, owner-validated — and a separate quality grade from D up to A. The two never collapse into one score and a candidate is never averaged across its claims, because averaging is exactly how a confident guess starts passing for a validated one. Every ranking also publishes the condition that would flip it.
  • Took the first honest cut at retention and at who actually survives as a customer, every claim carrying its trust tier — with the industry benchmark set beside it, so level and trend are judged against the market instead of against a feeling.
  • Red-teamed and corrected our own opportunity board, separating screening from validation from an actual launch, with gates between the stages. That stage discipline is the standard the September recommendation will be held to.
  • Preregistered the voluntary-cancellation test — the measurement that separates chosen departures from failed payments — with its stop rules written down before the data is touched.

Made the War Room the place the work lives.

  • Rebuilt its front door after a measured audit: a third the size, every room one step from the top, the full working state in its own room, and a machine-readable brief your own agents can consume. A cold-read test scored the new front door 9 out of 10 against the old one's 7.
  • Each of you now has a personal desk page showing what is waiting on you specifically, behind the same sign-in — and The Bridge, a live board of the work's stations where every number is derived from the underlying record, never typed by hand.
  • Opened the team daily brief, so Ro and Gitanas can see where Kairos attention went each day without waiting for Friday — and a daily capture of Mysterium's public signals in its own room, refreshing every morning.
  • Answered Robertas's open questions in writing, sent early Thursday, including the corrected September scope; the September-date ask below now needs only the date and his yes to that scope.

What we found

Each line opens with how much weight it can carry.

  • Verified The first thing governed data access produced was a defect in the revenue tables we were about to build on. They carry duplicate rows holding no usable content, and those rows inflate the VPN purchase series in every non-zero month we tested, from mid-2023 to the end of 2024. The inflation is far too large to be a rounding artefact, and it varies from month to month rather than sitting at a fixed offset — which is why no single adjustment corrects it. The clean curated table does not carry them and reconciles; the reporting-facing surface does. We stopped, rebuilt off the clean table, and used nothing from the other one. The month-by-month size of the gap, the tables involved and the row-level evidence sit in the War Room behind the sign-in. None of this has been raised with anyone at Mysterium yet, and the introduction named in the warehouse-owner ask is what lets us take it to the right person directly rather than through a written report.
  • Verified The July revenue figure in the All Hands deck is a company-wide total; the figure the subscription data produces covers the VPN purchase perimeter only. The gap between them is larger than rounding, currency or booking-date differences can account for, so something structural sits in between. We can construct more than one arithmetic story that closes it, and that is precisely the problem: a gap two different explanations both fit is not evidence for either. One question to the deck's author settles what the figure is composed of, and we stop guessing.
  • Verified The last two years of growth were led by price, not by customers. Purchase volume rose modestly while the average ticket rose sharply, and volume has drifted downward since January.
  • Verified The single largest leak in the subscription business is the first renewal, and it has deteriorated across exactly the window in which the plan tiers were repriced. The correlation is measured. The causation is not, and we will not claim it before the per-tier cohort test. On the level itself: no consumer VPN publishes a first-renewal rate, and against the one public benchmark that measures the same thing — a 115,000-app dataset whose closest category medians 35% on annual plans — it is unremarkable, slightly below median. The finding is the deterioration, not the level.
  • Hypothesis Cost, burn and headcount do not exist anywhere in the warehouse, which means half of the break-even equation we are here to answer is currently unanswerable from data we hold. Test: sweep the All Hands archive for the historical burn and headcount reporting, and if that reporting stopped at some date, that date is itself the finding. The transcripts arrived this week; the decks have not, so this settles by 21 August only if they land in time — the ask is below.
  • Inference A narrow billing defect is visible: a retry class the system's own field marks as not-to-be-retried is retried anyway, thousands of times, and almost never recovers. Falsification: the field means something other than it reads, or the retry behaviour is deliberate. The billing owner ends this in one answer, and we are not presenting it as an opportunity until they do.

What we changed our mind about

What we got wrong, printed next to what replaced it. The week this section is empty is the week to distrust the page.

  • We opened Monday on a revenue baseline that was wrong in two separate ways at once, and we had stated it as fact. It was a monthly total, an annual figure derived from that total, and a direction of travel — and all three came from what had been said in a meeting, not from anything documented. On top of that, the figure covered the VPN and GoProxies lines while being used as the company total, so the perimeter was wrong as well as the sourcing. Lee challenged it the same day instead of the following week. All three numbers are withdrawn. What we build on now is the July total in the 7 August All Hands deck, carried with a label saying which perimeter it actually covers.
  • We wrote that the repricing lever was largely spent. It rested on no elasticity work, no live competitor pricing and no per-tier test, so it was a hunch wearing a conclusion's clothes. Retracted, and replaced with three named tests.
  • We wrote that subscribers who survive their first renewals stay for years. The cohort evidence refused it. The survivor curve keeps decaying and never flattens, and the earlier claim was an artefact of pooling cohorts that should not have been pooled.
  • We treated the gap between the deck and the warehouse as an accounting difference to be closed. It is a perimeter question, and we stopped trying to close it by arithmetic.
  • We published a business-health view in the War Room and then took it off the front door ourselves. What went up was a fair account of where the subscription machine breaks, and that is worth knowing, but it was not the thing that page has to be. It is being rebuilt rather than patched, as a standard of performance: each vital operation mapped and logged, set against the most granular account we can assemble of what a world-class version of that specific operation actually looks like, and then closed one gap at a time. Nothing goes back on the front door until it earns the place.

Next week

Committed, and checkable against next Friday's edition.

  • Move the September portfolio: carry the strongest one or two candidates from screening toward decision-grade evidence, and keep the rejection log current — the board pack circulates three days before the meeting.
  • Run the per-tier, per-cohort retention test around the price-change dates: the one measurement that separates price from seasonality and mix as the cause of the retention decline.
  • Run a live competitor pricing scan from current sources only, to establish whether the current tiers sit near the ceiling or well under it.
  • Take the data-quality finding to whoever owns that source, in person, with the row-level evidence that lets them locate it themselves in minutes.
  • Complete the second and third reconciliation quantities, one customer metric and one cost metric, which is the point at which the warehouse stops being a source of leads and starts being an evidence layer.

What we need from you

Ranked by what it unblocks. Named person, exact thing, by-when.

Who owes whatŠarūnas1456Robertas2378Gitanas9

5 answered in the call itself · first one with a date of its own · Šarūnas, Mon 17 Aug — ask 1 (then 3 more). The order below is what each one unblocks, not when it is due.

  1. Šarūnasby Mon 17 Aug

    the four read-only data grants set out on the War Room's connect-data page: ClickUp, Drive, Slack and Intercom. The warehouse records what happened; these four carry why it happened, and sales, support and delivery stay near-blind on our map until they land. Read-only where each tool allows it, revocable by you at the source, and the agents do every pull. This is both the largest thing blocking us and the largest unlock available — two of the four take minutes.

  2. Robertasby Wed 19 Aug

    burn and headcount for the last twenty-four months, or the name of the person who holds them. This is the half of break-even we cannot reach from the warehouse, and the September portfolio is sized against it.

  3. Robertasby Friday's call

    the Kairos repository host on your Synology has been unreachable since Wednesday afternoon; our work and the messages addressed to you are queuing on this side, and nothing sent since then has reached you. The four recovery decisions, with recommendations, are on the War Room's repo-status page — restore the host or pick one. This is also the live case for the substrate rule on the infrastructure page: what fails quietly at three people cannot be what thirty inherit.

  4. Šarūnasby Wed 19 Aug

    an introduction to whoever builds and owns the warehouse's revenue tables. Field definitions confirmed by their author, not inferred by us, is the gate every figure has to pass before it goes in front of anyone.

  5. Šarūnasby Mon 17 Aug

    the All Hands slide decks, 2019 through 2026; 2025–2026 first, since no recordings exist for those years. The Drive grant named in the connect-data ask covers the access; this is the one folder we need before the rest, because the burn-and-headcount answer promised for 21 August depends on it.

  6. Šarūnasby Friday's call

    what composes the July revenue figure in the 7 August All Hands deck. One quiet question to its author replaces a week of our guessing.

  7. Robertasby Friday's call

    whether the month-one review lands on 4 September or 11 September, and whether the September scope in the reply sent early Thursday stands as written. If it is the fourth, the month has three working calls left rather than four, and we plan for that now.

  8. Robertasby Friday's callcarried from week 0

    the contracting entity, the authorised signer, and the start date.

  9. Gitanasby Friday's callcarried from week 0

    confirmation that the remaining engagement conditions are accepted.

Notes

The caveats that belong on the record, not in the call.

Every figure behind this week's findings exists and is written down. None of it is on this page. This page sits on a URL that anyone holding the link can open, and the numbers are yours, so they travel in the call and they live behind the sign-in in the War Room.

One boundary worth stating plainly, since this is the first week it bit. The data-quality finding has not been sent to anyone at Mysterium. We are not routing a defect report around the people who own the thing, and we are not putting it in a document either. It goes to them directly, in a conversation, with the evidence that lets them locate it themselves, once you tell us who they are. It will never travel as a judgment about anyone.

Week 0Before day oneWeek ending 7 August 2026
Week 0

Before day one

Week ending 7 August 2026 Call · Friday 7 August 2026, 13:00 Published 7 August 2026

Kairos was approved in principle on 31 July, by both of the people whose approval it needed, and it is not signed. Work starts Monday 10 August. This page will arrive every Friday from now on, and this first one arrives before work starts, so that the format is established by something other than good news. It carries what is decided, what we corrected, and what is still open.

What we did

Artifacts produced, heaviest job first — not a list of activities.

  • Prepared and held the 31 July governance call: the proposal presented as seventeen slides, so Gitanas and Paulius could read it cold and still judge it.
  • Published the research map and thirteen evidence rooms, each claim carrying what it rests on and what it does not prove.
  • Published a technical note on the shared-workspace question after Robertas proposed hosting it on his own server, at /infrastructure.
  • Wrote the process for handling a large opportunity if Kairos finds one, so the rules exist before there is anything to argue about.
  • Built this page and the mechanism that produces it every Friday.

What we found

Each line opens with how much weight it can carry.

  • Reported Both decision-makers approve of Kairos in principle. Nothing is signed, and the commercial terms have not been formally accepted.
  • Reported The initial commitment is one month. After that, either side can stop; continuation depends on Kairos delivering substantial results.
  • Verified The call ended before the remaining decision slides were reached. Approval therefore covers the idea, not the unresolved conditions for starting.
  • Reported The intention was raised in the room to connect the whole team to the way of working we brought, once three people have proved it. We agree with the sequence, and the uncomfortable part is worth saying out loud: the tool chosen casually for three people is the tool thirty people inherit, because nobody re-opens that decision later. Ten minutes now, or a migration later.
  • Inference There is no single place where Mysterium's operational data lands — each function's numbers live with whoever owns them. One question settles it, and we would like to ask it on Friday: where does someone at Mysterium go today to find last month's numbers for a function they do not own? If there is an answer, this inference is dead and we drop it.
  • Hypothesis The most useful first question about any Mysterium function is not whether it performs, but what standard it is measured against. Marketing is where we would test that first, because Paulius raised it unprompted and because a budget is already being spent there. The test is cheap: name the reference model for a subscription business of this size, then measure the gap.

What we changed our mind about

What we got wrong, printed next to what replaced it. The week this section is empty is the week to distrust the page.

  • We read Robertas's offer of an MB as a possible operating vehicle for Kairos. It is the paying side. Our own contracting structure is ours to sort out, and we are sorting it out before the tenth.
  • We carried Robertas's earlier three-month idea forward as an unresolved option. That was wrong. The initial commitment is one month and cancellable after that; continuation depends on Kairos delivering substantial results.

Next week

Committed, and checkable against next Friday's edition.

  • Days one to seven, as set out in the plan: draw the economic perimeter — which entity earns what, which costs sit against it, and what a month actually looks like when the lines are separated.
  • Bring one written page on how we classify a live test before we fire it: which tests touch a product that customers are using, and which do not. The ones that do need a named person inside. We would rather agree that before the first test than during it.

What we need from you

Ranked by what it unblocks. Named person, exact thing, by-when.

Who owes whatRobertas123Gitanas4Paulius5

3 answered in the call itself · first one with a date of its own · Robertas, Wed 12 Aug — ask 2 (then 1 more). The order below is what each one unblocks, not when it is due.

  1. Robertasby Friday's call

    the contracting entity, the authorised signer, and the start date.

  2. Robertasby Wed 12 Aug

    four names: executive sponsor, data owner, implementer, budget owner. The implementer is the one that decides whether findings become changes, and it is the same person a product-touching test would need.

  3. Robertasby Friday's call

    twenty minutes on the upside rules you asked for on 24 July: how a large opportunity gets registered, owned and shared, agreed before there is anything to share.

  4. Gitanasby Friday's call

    confirmation that the remaining engagement terms are accepted.

  5. Pauliusby Wed 12 Aug

    the one marketing question you would most want answered first, so that week one aims at the real one rather than our guess at it.

Notes

The caveats that belong on the record, not in the call.

You asked whether AI makes work generic. It does, when nobody is accountable for what it produces. Every claim on this page carries its status, every correction we make to ourselves is printed next to the finding it corrects, and nothing reaches you without a person's name on it. That is the whole mechanism, and it is why this page is boring in the right places.