Consolidated position · November 2026
This document brings product, market, audience, and marketing analysis together into one position: what already works today, what is being built on investment, and how each claim is backed — by a number, a source, or a direct experiment, not a declaration.
The same product is honestly described by two different promises — because it has two different layers: the method (already working) and distribution (not yet built). These are not two different products, and not a contradiction — they are two layers of the same system, and it makes sense to show an investor and a client exactly the layer that answers their question.
Arkhitsyn is a discipline in which every important decision in a project is first recorded with its alternatives and the reason for the choice, and then a separate, independent review looks for an error in it before that decision turns into code or a contract — so the record shows not only what was done, but also what was rejected, and why.
The mechanism already works and has been measured across 11 pilots, but today it is sold the one way that does not scale — personally, through hands-on guidance. The investment buys the missing half: certification, a partner network, a platform — and that half is what turns the economics from a single channel with unmeasured CAC into a funnel with two measurable channels.
Said about the same thing, the promises "already ready" and "not yet" would be incompatible. The contradiction disappears once the layers are named explicitly: the method — the cycle "alternatives → recorded decision → independent review → a journal that only grows" — already works and has been measured. The sales channel and the product-as-a-service are not built at all: there is not a single paying client outside the internal pilots, no certified partners, no separate platform with its own interface. The client is shown the first layer, the investor the second; both statements are about different parts of the same system, not the same part stated twice.
The single axis on which all four strands of analysis converge at once — product, market, audience, and marketing, not just two of the four. It is the combination of three mechanisms together: recording alternatives before a given option is scored, an independent reviewing party that must name a specific factual error in the decision, and a decision journal that only grows forward — an error in it is never erased by editing, only acknowledged by a new entry next to the old one.
Selection criteria and weights are fixed before numbers are plugged in — a safeguard against fitting the conclusion to a decision already made after the fact.
A separate reviewing role is required to name a specific factual defect in the decision before it moves forward — not to silently sign off.
The record of decisions and their reviews is never rewritten after the fact — an error is added as a new line, the history is never swept away.
Where six systems and arkhitsyn land on the map, plotted on the two axes of the decision protocol.
quadrantChart
x-axis "No alternatives" --> "Alternatives before scoring"
y-axis "No review" --> "Independent review"
quadrant-1 "Both present"
quadrant-2 "Review only"
quadrant-3 "Neither"
quadrant-4 "Record only"
arkhitsyn: [0.92, 0.9]
PRINCE2: [0.45, 0.55]
Scrum: [0.25, 0.35]
PMBOK 7: [0.3, 0.2]
Waterfall: [0.08, 0.08]
Kanban: [0.2, 0.08]
Agile Manifesto: [0.32, 0.14]
None of the six project-management systems below has all three mechanisms at once — each covers one or two, never the combination.
| System | Strength | What it lacks from the triad |
|---|---|---|
| Agile Manifesto | Top-level values, minimal bureaucracy | Prescribes neither a record of rejected alternatives, nor an independent review, nor a growing journal — a declaration, not a protocol |
| Scrum | Clear roles, retrospective on a regular cadence | Increment readiness is a quality bar, not an audit trail of rejected options; there is no separate role tasked with catching specific factual errors |
| Waterfall (as practiced) | Predictable sequence of phases | Architecturally forbids going back, instead of making rollback a measurable criterion — even the originator of the idea in 1970 included feedback loops that the practiced version later dropped |
| Kanban method | Flow management, work-in-progress limits | Manages how many tasks are in flight at once, not how a decision is made inside a single task — an orthogonal layer, not a substitute |
| PRINCE2 | Business-case review between project stages | The stage-boundary decision is made by a reporting committee, not by a formalized independent role with a "defect found → verified → fixed in the same record" protocol |
| PMBOK 7 | Flexible principles instead of a rigid process | The principles are not operationalized into a concrete format — who exactly records a rejected option, and in what file, is left to each organization |
| Notion / Trello (tools, not methodologies) | Recognizability, a ready-made interface, a flexible database or status board | Task status is not decision structure; neither one describes an independent reviewing role separate from the author of the page or card |
Notion and Trello are software tools with a concrete interface, not decision-process methodologies, so comparing them runs on a different axis than the six systems above: not "what it prescribes for the process," but "what the product gives the user out of the box."
By its own official description, Trello is a board where cards move between lists to reflect a task's stage; there is neither a record of rejected options nor an independent approving role in that. Notion is a flexible store of pages and databases that can be laid out any way, including as a "board" view; Notion's own documentation explicitly names the risk of information going stale and offers a reminder for the page owner to come back and check it manually — meaning the author and the reviewer are the same person here, not separate roles.
One line here does not favor Arkhitsyn: neither Trello nor Notion needs to explain what it is — both have millions of users and an established vocabulary of their own. That is a matter of brand recognition and user-base size, which no mechanism can quickly make up — a separate question from which mechanism is built into the process.
While this position paper was being assembled, the process itself twice acted on its own findings: when it turned out that a person's repeated intervention in work already underway was not being recorded verbatim, a new recording rule appeared; when decisions about the process itself were once made bypassing the independent review, all three were flagged as needing rework once they actually went through it — and since then, decisions about the process also go through it. For an investor, this is not a cautionary tale but an observable track record: the system being offered for sale, at the moment of the demonstration, runs exactly the mechanism it is selling.
works today limitation with a workaround growth potential — these three labels are used throughout the document instead of final verdicts.
Stated without professional jargon and verified by retelling — twice, in two independent runs.
The wording was retold in its own words twice, by an independent run each time, without peeking line-by-line at the original, and checked against two questions: does the retelling answer "what is this" and "why does it matter to me," and does understanding it require words like "agent" or "methodology" without explanation.
When a computer program helps make decisions on your project, there is a risk it gets something wrong — and you find out too late, once fixing it is already expensive. Arkhitsyn is a set procedure: before a decision turns into action, it is first written down — what was proposed and why exactly that way — and only then does a separate review, independent from the first, look for its weak spot. It's like a second medical opinion before surgery: not a replacement for the first specialist, but insurance against their blind spots, plus a written trail in case you later need to figure out what went wrong.
Neither run lost the meaning, and neither used unexplained jargon — the criterion is met. A fair boundary: both checks were run by the same class of tool in different sessions, not by a live person outside the topic — this is growth potential, a live test on a non-technical listener is still ahead.
The organic channel and the partner network are objectively different funnels, with different margins and different, differently-unknown customer-acquisition costs. A single averaged number would hide that difference rather than simplify it — hence the table by channel below, not a single line.
| Channel | Partner commission | Effective margin | LTV | CAC | LTV:CAC | Status |
|---|---|---|---|---|---|---|
| Organic (word of mouth, niche communities) | — | 65% | $32,500 | $3–8k | 6.5 (range 2.4–17.3) | works today |
| Referral partner (introduces the client, does not implement) | 10–15% one-time, year 1 | 57.5–60% | $28,750–30,000 | not measured | 5.8–6.0 | growth potential |
| Integrator partner (implements and supports the client) | 20–25% recurring on the "support" share (60% of ACV) | 50–53% | $25,000–26,500 | not measured | 5.0–5.3 | growth potential |
| Sensitivity: same integrator, if commission is taken on the full ACV | 20–25% | 40–45% | $20,000–22,500 | not measured | 4.0–4.5 | limitation with a workaround |
Industry benchmark for comparison: median B2B SaaS LTV:CAC ≈ 3.2–4.1:1, minimum viable threshold ≈ 3:1.
CAC in both partner rows is borrowed from the organic channel. The partner channel's acquisition cost is in principle different, likely lower, but no measured number exists for it anywhere — the rows read as "at the same acquisition cost," not as a forecast.
The lower bound is kept separate and not hidden. Under the worst combination of assumptions (ACV $15k, CAC $8k), organic gives 2.4, and the integrator with commission on the full ACV gives 1.5–1.7 — both below the industry threshold of 3:1. That is expected at a stage without a built-out sales funnel, not a failure of the economics — the point where a built channel makes CAC measurable and predictable is exactly what the investment buys.
The choice of base for the partner commission (the full ACV or only the "support" share) is an explicit decision, not a fact from a contract. The partner program does not yet exist, so the commission base has not been fixed with anyone — the difference between the two options shifts the resulting ratio by roughly a third, which is why the table shows both.
Market size is calculated top-down, with a source and a date at every step of the funnel — not a single point estimate.
works today
scales well, low marginal cost per client
works today
high margin, but limited by the number of people who can personally guide clients
growth potential
a distribution network without direct sales — not a single certified partner yet
growth potential
recurring revenue with better scale — today the package runs locally, not as hosting
The first and second columns work today — exactly what the client buys. The third and fourth are what the investment buys, and they are what opens the path to the upper bound of SOM.
The specific code and text of the package are protected by copyright automatically, without registration — that is a form of expression. The abstract process of running a project as a method is not protected by copyright anywhere (the law explicitly excludes methods and processes from that protection), and patenting a business methodology in the US is a separate, expensive, lengthy procedure with an uncertain outcome, not free protection by default. The brand name could be registered as a trademark — that has not been done yet, limitation with a workaround: nothing blocks current work, and registration is a simple step worth doing before the first contract. Today's real anchor is not a legal monopoly but the fact that the product is sold not only as code but as hands-on guidance, which is harder to copy than the text of the files.
Two buyer profiles, with KPIs quoted verbatim from 2026 industry surveys — not inferred.
In one of the pilots, the journal did not preserve the text of a person's repeated intervention in work already underway — after the fact it became impossible to establish whether the system raised the question itself or was told to, and that once led to a hasty conclusion that had to be retracted. The fix is a rule: a person's repeated intervention is recorded in the journal as a separate, verbatim entry, never reconstructed after the fact. This lands directly on Profile 1's "AI vs. human attribution" KPI (55% → 16% with mature practice) — not a declaration, but a concrete structural rule already built into the process.
A decision made in conversation, without recording the alternatives, once turned out to be incomplete and cost roughly an hour of wasted compute time. A heavier case of the same pattern: three decisions about the process itself were made bypassing the independent review, and when they were actually run through it, all three were flagged as needing rework, one of them containing a defect that made the mechanism not work for most of the roles involved. The fix is a rule: decisions about the process itself must also go through independent review before being handed off, no exceptions. This lands directly on Profile 1's governance concern (36%) and on the single most common concern about quality and oversight of AI output.
While working on two pilots in parallel, the process produced dozens of real, technically verified findings — but almost all of them stayed inside individual work cards, never becoming reusable knowledge, until a person pointed this out directly. The fix is a rule: everything verified and accepted as a basis for work must land in a separate knowledge base, not only in the card for that one task. This lands directly on Profile 2's pain — 31% of a developer's time going into invisible work, including re-discovering something already known. growth potential: the rule was added very recently, and a second independent case confirming it holds up in practice has not yet occurred.
Three channels, chosen not arbitrarily but based on where the audience already physically is.
Developer communities around AI-agentic tools together number from hundreds of thousands to over a million members — an exact target profile, not a random ad placement. works today as the cheapest measurable step.
Closed or semi-open industry summits draw decision-makers specifically, not just practitioners — an audience already filtered by the organizers. growth potential: entry is paid and competitive, and an application without a ready case study risks rejection.
An existing network of over 150,000 certified consultants and tens of thousands of registered firms — a ready-made, pre-screened pool of clients already paying for AI-agent implementation. growth potential: a high entry bar, and the network is already large and competitive for attention.
The tiering is built on benchmarks from two adjacent industries — consulting referral fees (typically 5–15%) and SaaS partner programs (median 20–30%) — not on observing an actual partner program of our own, which does not yet exist. The network around Claude was used as evidence of audience size and profile ("where it physically is"), not as a source for the commission figures themselves — its internal economics are not publicly disclosed. limitation with a workaround: with no real partner yet, the tiering is the best available estimate, not an observation; the workaround is to lock the commission base into the contract of the first partner program and recalculate from actuals.
How long it takes to go from an empty project folder to the first confirmed working cycle — measured by timestamps, not eyeballed.
This is one measured data point, not a statistical sample — another task could turn out easier or harder. The documentation storage structure itself creates no barrier (it is an ordinary folder scheme for managing personal projects, already familiar to people) — the barrier comes from the layer of roles and independent reviews on top of it, and that layer is exactly what took the measured 4 hours 10 minutes. growth potential: a second full measurement on a different task has not yet been taken, and a quantitative comparison against fully manual work on the same folder framework without the roles also remains an open step.