What You Can Genuinely Ship

The honest map: more than the doubters say, less than the demo implied — and the boundary is supervision.

More Than the Doubters Say

Time to say plainly what this course believes you can do, because it is more than the doubters say. A non-engineer with AI can ship a real product for a real small business, and run it responsibly for years. The shape of that product is familiar by now: a bounded set of users — two hundred customers, not two hundred thousand; bounded data you have inventoried and could look someone in the eye about; actions that can be undone; and a person in the loop wherever a mistake would touch somebody. The groomer's booking page is exactly this shape, and so are a striking number of genuinely useful products: the club rota, the tuition scheduler, the repair-shop job board. Five years ago, every one of them needed an engineer or went unbuilt — and mostly they went unbuilt. That is the change. Not that building became trivial, but that this whole shape of product became yours to make.

A real product for a real small business, run responsibly for years
The shape: bounded users, bounded data, reversible actions, a person in the loop
Rotas, schedulers, booking pages — products that mostly went unbuilt before
The change is not that building got trivial, but that this shape became yours

Less Than the Demo Implied

Honesty has to cut both ways, so here is the other edge. Some products you still cannot responsibly ship, and the demo's confidence does not change that. Products where mistakes cannot be undone, or where the law has strong opinions: moving money, health, anything touching children's data. Products that act on their own at a scale no person could review. And — the most useful test of all — anything you could not afford to watch, because you now know that running a product means watching it, answering for it, and putting it back when it breaks. Notice where the boundary sits. It is not intelligence, yours or the AI's. It is supervision: can a person see what the product does, catch it, and reverse it? Vibecoding's final module drew this line for prototypes. It holds for products too, and no demo, however dazzling, has ever moved it.

Irreversible or regulated ground — money movement, health, children's data — is out
So is unsupervised action at a scale no person could review
The plainest test: could you afford to watch it, answer for it, reverse it?
The boundary is supervision, not intelligence — and demos do not move it

The Quiet Dignity of the Small Real Thing

Put the two halves together and you get the course's closing picture. A booking page that one business relies on every working day, for years — backed up, watched, supported, ready to hand over — is a real product in every sense that matters. It beats the Gmail rebuild, because the Gmail rebuild was never going to happen — and not because you aren't clever enough. Gmail is spam filtering at planetary scale, twenty years of edge cases, deliverability, compliance, and a support organisation; it was never going to be anyone's clever weekend. The delusion this course exists to correct was never "I can build things". That one is true, and you have the working product to prove it. It was "the demo was most of the work". It never was: two days of twenty-three. Once you see the ratio clearly, you can stop being disappointed by it and start planning with it.

A booking page one business relies on for years is a real product
It beats the Gmail rebuild, because that rebuild was never going to happen
"I can build things" is true; "the demo was most of the work" never was
Seen clearly, the ratio is not a disappointment — it is a plan

Where to Go From Here

The shelf from here is short and good. Vibecoding is behind you: it taught the weekend, and this course taught the other twenty-one days. If your product has AI inside it, 'Does Your AI Actually Work?' is the course for testing that properly, with evidence rather than vibes. When the day comes that an engineer joins you, 'Securing AI Systems' speaks their language. And two books have outlived every tool they mention. The Mythical Man-Month, from 1975, on why adding people to late software makes it later — still true, because it was never really about software. The Phoenix Project, a novel about why shipping small and often beats the big terrifying release — you lived its argument in Module 4. Now, one last time, the exercise: your own project, all fifteen dimensions, and then the decision this whole course has been preparing you to make honestly.

'Does Your AI Actually Work?' when your product includes AI
'Securing AI Systems' for the day an engineer joins you
The Mythical Man-Month: adding people to late software makes it later — still true
The Phoenix Project: ship small and often — you lived this in Module 4

◆ Try it yourself

Fill in the fifteen-dimension comparison for your own project — the honest state of each dimension today, even where the honest answer is "nothing yet". Then make the call: ship, don't ship yet, or ship with help — and name the three dimensions that decided it.

The fifteen dimensions — my project, honestly, today:
1. Where the data lives: [ ]
2. What happens when data is wrong: [ ]
3. Backups you have actually restored: [ ]
4. Accounts and resets: [ ]
5. Who can see what: [ ]
6. Email that actually arrives: [ ]
7. Taking payment: [ ]
8. Getting changes to users: [ ]
9. Undoing a bad change: [ ]
10. Knowing it broke: [ ]
11. Cost at real usage: [ ]
12. Support — who answers: [ ]
13. Updates and rot: [ ]
14. Handover — could anyone else run it: [ ]
15. The second year — upkeep versus new features: [ ]

My decision: [ship / don't ship yet / ship with help]
The three dimensions that decided it:
1. [ ]
2. [ ]
3. [ ]

How you'll know it worked

Every dimension has an honest answer, even where the answer is "nothing yet"
Your decision is one of the three, written down — not "probably fine"
The three deciding dimensions would survive being read back by a sceptical friend

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.