Eight decisions turn your idea into a business. One more puts it live.

The rule the machine runs on, printed inside the product on the market panel:

“Every quantitative figure carries a source; any unsourced number is flagged ‘estimate · unsourced,’ never invented.”

390.5 seconds, measured, on the production system, on 17 August 2026. 403.4 including signup. Nine gates offered, nine accepted, nine of nine passing their value check.

I am calling that seven minutes on this page. It was six and a half. I rounded up, against myself — a company that rounds its own numbers in its own favour has just told you exactly how much to trust the next one.

And the honest sample size: that is one run. Not an average, not a benchmark, not "how long it takes." One recorded run on one idea, published so you can check it. Yours will be its own number.

Nine gates in 390.5 seconds is about 43 seconds a gate, and the generating is most of that. Nobody read an eleven-section business plan in what was left over. Let me say what that is instead of hiding it: that is a script clicking Accept.

You will read. You should. Reading eight documents about your own business properly is an hour, and it is the most valuable hour you will spend this month.

So here is the split, said now rather than found out at minute four: the software's seven minutes is the typing. Your hour is the deciding. Nothing advances until you accept it. Nothing is built on top of something you have not seen. And if a gate is wrong you do not start over — you type one line saying what is wrong and only that gate rebuilds.

You are not buying a machine that does your thinking. You are buying back the typing and keeping the thinking.

There is a stack of homework a real business does before it opens, and you skipped it. Not out of laziness — each piece is a week, you did not have nine weeks, and skipping felt like speed.

It was not speed. It was deciding not to find out.

Here is what came back in our run, with the specific details rather than the adjectives, because the details are the argument.

The idea, scored — and scored critically. 82 out of 100, weighted across four axes with every component bar exposed so you can argue with the arithmetic. The SWOT came back with four weaknesses and four threats, unprompted — "Customer acquisition cost could be high", "Possible user distrust of AI for pet training", "Limited appeal beyond first-time dog owners". It is not there to flatter you.

The market, sized, with sources you can click. A researched avatar that names what they do now, what it costs them, what they want, what keeps them up at night, what they quietly doubt about themselves and who they blame. Then TAM, SAM and SOM — each figure carrying two named sources and a stated methodology (top-down, CAGR 7%).

An eleven-section business plan. Problem, segments, UVP, solution, channels, revenue streams, cost structure, key metrics, unfair advantage, milestones, GTM. And then the financial model, already populated from the plan — $25.00/mo ARPU, 80% gross margin, a 500,000-person beachhead, every value labelled est. You arrive at a filled-in model to argue with. Never a blank form. And the executive summary is derived last, read-only, from the sections you actually accepted, so it can never describe a plan you rejected.

A brand that checks its own work. The palette came back with contrast ratios computed against the background — 17.4:1, 4.87:1, 4.8:1, 4.75:1 — and the line "All text-on-background pairs pass WCAG AA." Not "here are some nice colours." Colours with the accessibility arithmetic already done.

Messaging scored cell by cell. The matrix queued 15 cells of type × channel and six came back generated, each voice-scored 0–100 against the brand you just accepted — a 98 on the tagline, an 89 on the weakest. The other nine were still sitting in the queue when we accepted the gate, and the panel prints that on its own face — IN QUEUE 15, APPROVED 0 — rather than rounding them into the total. Two earlier drafts of this page called all 15 voice-scored; they were counting the queue, and the panel's own artifact — the messaging panel of the 17 August run, published in full at /the-run — is what corrected them. Approve is fail-closed below 70: a low score needs an explicit override. And when you advance the brand, every message pinned to the old version goes Stale and says so, rather than quietly going off-key.

Three priced offers with a computed stack. $15 / $25 / $40, three bonuses broken out at $50 / $40 / $35 against the $25 price — and the ratio the panel prints is computed from those bonuses, not chosen to look good.

A campaign plan that admits what it does not have. A $20,000 budget across three channels, plan score 85, three scenario projections, and paid search returned PAID — HELD: "Paid channels are planned-but-held — launch is held pending Budget-Approver authorization + the money-movement gate. They will not arm silently." And across the top of the plan, printed by the product itself:

> "Plan under-filled. Your co-founder's draft is missing start/end dates. This is an honest gap — nothing was fabricated to fill it."

Read that twice. It left a hole and told you, rather than filling it with something plausible. Every other claim on this page depends on you believing that sentence is real, which is why the whole run is published.

We have no customers and therefore no testimonials, and I am not going to invent any. What we have is one recorded run, published whole — including the parts that did not work.

Two of the panels in that run produced nothing. The competitor scan printed "No competitor scan yet". Positioning printed nothing either, because it derives from "competitor gap × the avatar's top pains. Regenerate to fill."

That is why competitor analysis is not on the list of what you get, on this page or anywhere else on this property. It is a gap. We are not going to sell it as a feature until it fills.

Two more, in the same spirit:

• The payment leg was self-signed. Our run signed its own webhook, so that leg is asserted, not observed, and our own ledger demotes it to PARTIAL rather than PROVEN. We are not telling you a demo took a real payment. What we are telling you is that the checkout is provisioned and chargeable on publish, which is a different and smaller claim, and it is the true one.

• Two of ten adversarial buyer attacks came back "blocked", not "defended" — the test harness could not set them up. That is a gap in the test, not a pass, and the run's own rollup records it that way.

I am publishing that because in about seven minutes you will be able to check every claim on this page on your own screen, and I would rather you heard about the empty panel from me than found it yourself at minute two.

The full run — every artifact, every timestamp, every finding →

The receipt itself: docs/_demos/runs/run-2026-08-17T13-44-19-657Z/a01-solo-petcoach/report.json.

Here is our arithmetic, all of it.

That is the whole model of the $7. It is not a discount, it is not a loss leader with a hook in it, and it is not credited against anything later. It bought your build. If we credited it we would have to find that money somewhere else on your bill, and I would rather charge you seven dollars once and tell you plainly what it was for.

And here is the table you will not find on this page: what an agency, a strategist and a copywriter would have charged you for the same eight documents. Every version of this page before this one had that table on it, running into five figures. I took it out. We have never charged anyone five figures for this, nobody quoted us those numbers, and printing them would mean inventing a figure to make our own price look good — thirty seconds after telling you the machine flags any number it cannot source.

If you want to know what the alternative costs, go and get a quote. That number will be real. Ours would not have been.

Publishing, and your assets in anything but a PDF. Two lines, both of them here rather than at the last screen.

You keep, permanently: your entire run, readable in your account, and a PDF of every artifact — your brand's colours and type applied to the document. Read it, print it, hand it to your accountant. Come back in six months and it is still there. The PDF is never taken away: not before a subscription, not after a refund, not ever.

What needs a plan is DOCX, GOOGLE DOCS, MARKDOWN, JSON. Those are the editable, re-usable files rather than the answer, and asking for one without a live plan returns a plain plan_required and a link, not a broken download. If what you want is to read what your $7 built and keep it, the PDF is the whole of that and it costs nothing more.

And Publish is the other line. Going live — a real address, pages that serve, a checkout that can take a card — is where the subscription starts. The refusal says so in the product's own words: "Publishing needs a live plan. Your run is saved — pick a plan and publish it whenever you're ready."

You can come back and publish any time. There is no window, no expiry and no penalty for leaving it in the drawer. Every read, every re-open and every PDF stays open to you whether you subscribe or not.

And two sentences I would rather you read here than discover on the third screen:

• There is no free trial. Your first payment is a full first month at the full price.

• The $7 is not credited against it.

Deliberately, and the reasoning is short.

A trial is a fortnight in which you do not have time to evaluate anything, ending in a charge you had forgotten about. A free tier is either a crippled product or a surprise later.

What you actually want is not free software. It is proof before you pay. So you get the proof — the whole product, on your own idea, start to finish, for seven dollars.

By the time anyone asks you for $79, you have run every gate the $7 buys and you are holding the eight documents it produced. There is nothing a trial could still tell you.

Onomaco has 26 apps in it. You will not set up 26 apps.

We are not going to list them here, because the list is the problem. Every app you name raises your estimate of how long setup will take — and setup is the documented reason people quit platforms this broad.

Everybody in this category thinks the product is capability. It is not — capability is table stakes, and they all have it. The product is assembly, and every one of them leaves assembly to you.

“4 months into using this program and still getting things configured.”

“I had to hire consultant to teach me.”

“Every tool was good. None of them talked to each other.”

Public reviews of platforms in this category. No competitor is named anywhere on this site — the counsel flag on quoting them unattributed is still open.

We took that job. The co-founder reads your business, picks the handful of apps it actually needs, wires them around one customer record, and publishes.

And when it goes live it runs as one sequence, in this order: your offer becomes chargeable · your funnel pages publish · your campaign arms · CRM intake opens. If any step fails the whole launch rolls back. Nothing is left half-live.

At the launch screen the product stops and lists what it cannot do on your behalf. In our run that list read:

• Stripe — "Stripe is not connected — needed to get paid."
• Email (Resend) — "Connected — used to send launch + nurture email."
• Social (Meta) — "not connected — needed to publish the launch campaign."
• Calendar — "not connected — needed to take bookings."

It does not pretend to have connected your Stripe account. It asks you to.

And one more thing, which is a property of the code rather than a policy in a document: money movement is permanently human-gated. Affiliate payouts are never sent automatically — the agent dispatcher refuses them outright and hands them back as a manual step. Software that moves your money without asking is not a convenience.

Seven dollars for the run. A subscription if you publish. Metered AI beyond your plan's allowance. That is the list.

Your buyers pay into your own Stripe account, and we take no percentage of it. No plan on this site carries a revenue share. We do not charge for support, and there is no partner tier that costs extra to teach you the thing you already bought.

Your plan includes an AI-credit allowance and the product tells you in-app when you have used 75% and 90% of it. Past the allowance, credits cost the published rate — $0.03 a credit, the same rate on every plan — and because a heavy generation draws 3 credits, a researched market pass past your allowance is $0.09, not $0.03. The whole arrangement, including the part where you pay us more, is on the pricing page.

Seven days, full refund, no questions. It covers the $7 like it covers everything else bought through an Onomaco checkout — ask within seven days and we return the full amount. You do not have to finish the run, reach a result, or give a reason. The refund policy governs, and there is no offer-specific exception to it anywhere on this site.

What is not built is the automatic one, and every earlier draft of this page carried it: if the build doesn't finish, you don't pay — refunded by itself, no ticket, no email, no asking.

There is no automatic refund in this product today. A refund you have to ask for is a smaller promise than one that fires on its own, and I would rather name the gap than let the same words cover both. So the sentence above is the one that binds us, and the automatic version goes up the day it fires by itself — the same rule that keeps competitor analysis off every list of what you get.

You did not avoid launching because you are lazy. You avoided it because doing it properly was nine weeks of typing and you never had nine weeks. So the choice was always: do it badly, or don't do it.

There is a third one now. It costs seven dollars, and in our measured run it took 390.5 seconds.