Hard-to-copy asset
Household exits across a live residential network.
This is the source of possible differentiated access. It is not yet proof of customer demand, value or defensibility.
01 · the bet
Own the household route. Earn the right to call it value.
An unusual supply asset becomes a strategy only when it solves a valuable customer problem better than ordinary VPN infrastructure.
Decision stateTest before pursuitNot a candidate bet yet.
In plain words
Mysterium already routes traffic through real household internet connections. That is difficult for a conventional VPN to copy. We still have to prove that lawful customers need it, notice the difference, and will pay enough to support the cost and risk.
Concrete example: a destination challenges a normal VPN server but accepts a verified household exit in the same country, and the customer completes the task without losing protection.
Copyable product floor
Working Protection.
Truthful state, fail-closed traffic, usable data-path checks and bounded recovery. Mandatory even if Real Access dies.
02 · the product
The task works. The protection is true.
One customer promise, three honest states and one deterministic protection loop.
Promise stateProposedNo customer-facing promise is approved.
In plain words
The customer is not buying a tunnel icon, decentralisation story or server count. The proposed outcome is: my ordinary web task completes from an address the destination treats as a home connection, and I can trust that I remain protected while it does.
Concrete example: if the route stops carrying data, the app changes state immediately, contains unsafe traffic, tries only an equivalent repair and proves the repaired route before saying Protected again.
Protected and working
Tunnel, public route, DNS and usable data path agree.
Repairing
The unsafe path is contained while an equivalent route is tested.
Unprotected
No equivalent repair exists. The app stops and explains the exact condition.
Eligible attempt→
Path verified→
Degradation detected→
Unsafe fallback contained→
Equivalent repair verified
Machine boundary: an LLM may translate intent, explain the failure and assemble a support packet. It never decides whether traffic may leak.
03 · what the evidence carries
The network is real. The customer value is not established.
Supply evidence can justify a decisive test. It cannot substitute for that test.
Evidence ceilingAsset-level asymmetryNo causal demand or retained-value proof.
In plain words
The strongest numbers describe the residential network and how the current product is used. They do not tell us why customers chose it, which lawful task required it, or whether the route improved retention or contribution.
Concrete example: roughly 90% of paying connects use residential routes, but Plus includes residential access by design. Entitlement can create the observed share without residential being the purchase motive.
Counterevidence that stays visible +
- The only attempted direct job comparison found no destination that rejected the datacenter route while accepting the direct line.
- A 78-second peer blackhole left the client Connected with zero traffic, no recovery and no explanation.
- Provider death failed open without a kill switch.
- A crude error-to-churn association fell from 8.52× to 1.18× after observed-day and event-opportunity controls.
- One route resolved through a datacenter relay, one test slowed roughly sixfold, and one location label was wrong.
04 · the customer question
Do not invent an ICP around an asset.
Test lawful job families and operator-like behavior separately; choose a market only after the jobs appear unaided.
Customer stateNo ICP selectedC12 remains an eleven-way evidence co-tie.
In plain words
The current record contains plausible customer jobs, not a winning customer. The first research samples must preserve those alternatives and keep external-process behavior out of the value estimate.
Concrete example: a traveller completing a lawful home-country account task and a user rotating household IPs for an external process may create similar telemetry. They are not the same customer and cannot share one value claim.
Lawful test strata
J14 · J17 · J12 · J04
Residential-dependent tasks, challenge avoidance, cross-border continuity and ordinary restricted-network access.
Containment strata
I01 · I02
External-process and proxy-like patterns. Exclude from the value estimate and monitor separately.
Sampling frame
At most four installs
Use home country only as a stratifier. Neither rule creates an ICP.
05 · business model and go-to-market
Prove value before buying distribution.
Subscription economics and customer acquisition are separate questions with separate evidence.
Commercial stateUnpriced hypothesisNo new tier, spend or campaign is proposed.
In plain words
Keep the current monthly and long-plan structure while learning. First prove a lawful residential-only outcome and willingness to pay at the truthful current annual-equivalent price. Only then test an outcome-led offer with measurable traffic.
Concrete example: a residential-labelled offer recorded 134 purchases from 4,719 sessions versus 3,402 from 282,634 sessions across /offer. The rate difference is a lead only because traffic selection, spend and retained value are missing.
Business model
Residential success → specific willingness to pay → retained contribution.
Measure 90-day retained cash per eligible account after refunds, fees and traffic. Keep protection repair, lean operations and media gains in separate ledgers.
Go-to-market
One checkable outcome. One direct offer. One measured channel.
Do not lead with ideology, decentralisation or server count. Separate web purchases from app-store outcomes and hold out a concurrent comparator.
06 · the smallest decisive programme
Four gates. Each can stop the story.
Association, asset advantage, lawful job and mature retained contribution remain separate claims.
Execution stateDesigned, not launchedEach stage needs its own authority.
In plain words
Start with the cheapest tests that can destroy the idea. Do not contact customers or change the product merely because the descriptive data looks promising.
Concrete example: if a twenty-destination battery cannot find a repeatable lawful case where residential beats both datacenter comparators, stop calling this the One Big Thing. Keep fixing Working Protection.
- 0aDescribe the association
Within mature Plus accounts, join exit class to 90/180/365-day retained cash. This cannot promote the opportunity by itself.
- 0bTest the asset
Direct → Mysterium datacenter → Mysterium residential → managed-VPN datacenter. At least 20 destinations, three runs and two geographies.
- 01Find the job and willingness to pay
Run a consented unaided-job test across the lawful and containment strata at the truthful €40 annual equivalent.
- 02Test retained contribution
Only under a signed Opportunity Schedule, randomize eligible new Plus purchasers to a residential-verified experience versus a concurrent current-product comparator, then read 90-day net retained cash before 180-day follow-up.
07 · kill rules
The bet dies when the advantage, job or economics fail.
The product floor survives. The One Big Thing framing does not.
Decision disciplinePredeclared exitsMomentum cannot rewrite them after the result.
In plain words
A real strategy says what result would make it wrong. These three failures stop the residential thesis even if the network is impressive or the product repair works.
Concrete example: a safer, more reliable VPN is still worth shipping after a failed residential test. It is a better product—not evidence that household access is the company-shaped bet.
No measured access advantage
Residential fails to beat the better datacenter comparator by at least 15 percentage points on the datacenter-failing subset—or a bona fide search finds no such subset.
No lawful job or willingness to pay
At least 30% describe an external process while at most 10% name a lawful residential-dependent job; or definite yes is at most 10% and definite no at least 70%.
No mature retained contribution
The 90-day net retained-cash lower bound does not beat the concurrent comparator, or any customer-harm guard trips. If the required sample exceeds one quarter of eligible purchases, retire the framing as infeasible. A 30-day refund rate above 1.5× the 5.08% baseline is a hard harm signal.
08 · authority and record
Register the opportunity before leverage appears. Pursue nothing material yet.
The governance architecture exists to prevent success from becoming a memory and ownership dispute.
Hard gateNo signed scheduleNo material pursuit.
In plain words
This private release records the opportunity and its disproof programme. It does not authorize the programme. Cheap internal preparation may continue; any material build, launch, outreach, transfer or spend waits for the governing agreements and an opportunity-specific signed schedule.
Concrete example: the Stage 0b destination battery can be specified internally. Running a product-touching or externally recruited version still requires its separate authority.
Services Agreement
Ordinary work, data, confidentiality, IP, expenses and termination.
Mammoth Framework
How exceptional opportunities are recognized, registered, protected and measured.
Opportunity Schedule
Specific economics, rights, baseline, owner and signer for Real Access.
Lee–Šaras Agreement
Internal authority, allocation, departure and dispute rules before the first opportunity.
Review and source record +
- The recommendation-blind Tris reconstruction selected the household-exit asset as the company spine; it changed the strategy rather than certifying it.
- The v0.1.0 release review found five material defects. v0.1.1 and v0.1.2 preserve the first bounded repair cycle. Tris rejected v0.1.2 after finding sibling motion-evidence and clarity-denominator defects. v0.1.3 preserves the next failed review; v0.1.4 is its single bounded repair cycle. The review receipts live outside these immutable page bytes.
- Primary source: Kairos — One Big Thing VPN Opportunity, 3 September 2026.
- Supporting records: opportunity re-rank, VPN product immersion, retention economics, cancellation reasons, provisional ICP decision, jobs register and customer-journey audits.