The first 30 days

Nineteen lessons. You have researched a market, positioned a product, made a page, written a brief, found queries, drafted posts, designed a sequence, judged prospects and threads, and worked out how you would measure any of it. Today those stop being nine separate exercises and become one thing somebody could execute. From Monday, your plans become testable.

1. A plan you cannot be wrong about is not a plan.

Most marketing plans are lists of intentions. They cannot fail, because nothing in them says what success or failure looks like — so they quietly continue for a year.

A real plan makes a claim: if we do this, that should happen, by then. Which means it can be checked, and it can be wrong.

You have been building toward this since Day 9, when the task asked what would change your mind. Same idea, now across a month of work.

2. Two channels, done properly, beats eight done thinly.

You have now seen eight. The instinct is to use them all, and it is the most common way small teams achieve nothing.

Thin presence everywhere produces no compounding in any of them: a blog with four posts, a feed that stops in week three, an outreach list nobody followed up. Each looks like activity. None reaches the point where it starts working.

Two channels for thirty days, actually sustained, tells you something real. Eight tells you only that you were busy.

The two you chose on Day 9 are the ones. If the last two weeks changed your mind, change them — but change them deliberately, and say why.

3. Sequence, do not schedule.

A calendar says what happens on which day. A sequence says what has to be true before the next thing can work.

Week one and week four should not look alike. Early weeks are setup and small tests; later weeks are volume on whatever survived. A plan where week one and week four are identical has not thought about the middle.

	Typically
Week 1	Set up, publish the minimum, get the first real signal
Week 2	Fix what the first week exposed. Keep going.
Week 3	Increase what worked, cut what did not
Week 4	Enough evidence to decide whether the channel continues

4. Dependencies are what make it operational.

This is the difference between a plan and a wish, and it is where most fall apart in week two.

Write these into the plan explicitly. Every one of them is a thing that, done in the wrong order, wastes the work either side of it — and every one of them gets done in the wrong order by people in a hurry.

5. The minimum viable version of each channel.

For each channel, the question is not what good looks like. It is: what is the smallest version that can run in week one?

Not because ambition is bad, but because a plan requiring a team you do not have is a plan that stops on day four. And a small version that runs teaches you more than a large version that does not.

Then expand only what survived contact with reality.

6. Name who does each thing — including which agent.

Every activity in the plan has an owner, and after eighteen days you can be specific about which of those owners is a person and which is an agent.

Work	Agent drafts	Human decides
Research	Scout	What it means
Page copy	Herald	What we may claim
Articles	LotsBlog	The brief, and the thing no competitor has
Search	Ranker	Which queries are worth it
Social	LotsSocial	What is worth saying
Email	Courier	Whether the email should exist
Outreach	Envoy	Who to contact and who to exclude
Community	Ambassador	Whether we belong in that thread
Numbers	Gauge	What to do about them
Review	Editor	Ship or not

Notice the right-hand column. Every single one is a judgement, and not one of them has moved. That is what AI-native marketing actually looks like from the inside — the drafting moved, the deciding did not.

7. Say what you are deliberately not doing.

Day 9 taught that a recommendation without rejections is not a recommendation. The same applies to a plan.

State plainly: which channels are postponed and until when · what you will not make · which assumptions you are testing rather than asserting.

The value is not tidiness. It is that in week three somebody will suggest starting a newsletter, and a plan that already says "not until we have 100 subscribers from the two channels we chose" answers that in one line instead of an afternoon.

8. One number a week, and the condition for stopping.

From yesterday: each channel gets one number, on the rung it can honestly be judged on this early, with its window and denominator.

And each gets a stop rule you would actually honour. "If by 20 October we have not had a single reply that mentions the specific problem, outreach stops and that time goes to community."

Write it with a date. A stop rule without a date is never triggered, because next week always looks like it might be different.

9. What happens on Monday.

This is the last lesson. From Monday the work is the work — real tasks on LotsSocial, and the lessons stop being the main event.

A few things I want to say plainly before that.

Some of your plans will be better than mine. That is not encouragement; it is the reason the tasks were real from Day 10. I have been reading them, and the best of what comes in tomorrow is what we run.

You will each be stronger in different places. I have seen who catches errors, who writes, who judges fit honestly, who is rigorous with evidence. From next week that starts to matter for what you work on.

And the thing I most want you to take. Nineteen days ago most of you thought AI was a thing you typed into that produced text. What you have now is narrower and far more useful: these tools draft, and you decide. Every agent you met refuses to do something — Scout will not act, Herald will not invent a claim, Editor will not review its own work, Ambassador will not pretend it submitted a listing, Envoy will not send.

Those refusals are where the judgement lives, and the judgement is yours. That is the whole program.

Monday — the work starts. No lesson. Details in #announcements over the weekend, including who is working on what.