Thirty days ago I'd never heard of half of this

Thirty days ago I had a domain name and a rough idea.

Right now there’s a website I never wrote a line of. Six things on YouTube. Six of these emails. An AI version of me who insults me in public. And a system that emails people while I’m asleep.

Revenue: nothing. Not a dollar.

I was going to write about the broken bits again this week. I’ve done that four weeks running and I’m getting good at it. Then I sat down and made a list of what’s actually standing, and it stopped me — because thirty days ago I couldn’t have named half of it, let alone built it.

What’s actually there.

A website. A real one — a proper content system, not a page with a box on it. My entire contribution was three things: I paid for the platform, I said what I wanted, and I dragged some files into a folder. Not one line of code, not one line of copy written by hand. The only job the AI genuinely couldn’t do was drag the file, because it can’t reach my computer. That’s the whole division of labour.

ManyChat. I’d never heard of it. Not “heard of it and never got round to it” — never heard of it. It was suggested, I described the problem, and there is now a live automation system running on ManyChat. That link pays me a commission if you sign up through it, and your first month is free — more on why those links exist further down.

Buffer. Same. Never heard of it. It now schedules everything I post and saves me hours a week. No link for this one: I applied to their programme on the 18th of August and haven’t heard back.

This newsletter. One a week since Launch Day, no misses. It runs on beehiiv, which I picked before there was any commission in it and would recommend anyway. That’s also an affiliate link.

Six things on YouTube. Three proper videos — five minutes, four minutes, a minute forty-five — and three Shorts. Thirty days ago that number was nought, and I’d have told you confidently that I wasn’t the sort of person who makes videos.

A recurring character. He’s about ten months old, wears my glasses, has the manner of a disappointed accountant and calls me “old fella”. He’s an AI. He is also, by some distance, the most popular thing I’ve made.

A way of filming that works. Video one was one seventy-second piece to camera and it took eight takes. Video two was nine short beats, none over about thirty seconds, and eight of the nine went in one take. I hadn’t been bad on camera. I’d been attempting something actors don’t attempt either, which is precisely why films are shot in short takes. Three things changed at once — the beats, a teleprompter, and it being my second time — so I can’t honestly tell you which did the work. I’d bet on the beats. The teleprompter was Teleprompter.com; that’s an affiliate link, I’d earn a commission if you signed up, and I’m only mentioning it because I actually used it.

And TikTok, which I’d said I wasn’t doing. It was on a written list of things I would not touch for ninety days. It felt like a distraction, and if I’m honest it felt like somewhere I’d be too old to be standing.

Three posts in, it has ninety-eight followers. Three posts. I’ve done nothing clever to earn them and I wasn’t really trying.

I haven’t got an explanation and I’m not going to invent one — three posts isn’t a strategy, it’s an accident I should probably pay attention to. Which is the whole lesson of this email in miniature: I was wrong about the platform, I don’t yet know why, and I’m going to keep going until I do.

A studio, sort of. I spent two days of my life researching studio lights. I do not have a studio. I have a room. I also already own a light, bought in July, which has spent most of the time since in a corner. Instead of buying anything I worked out how to switch on a virtual background. Same room, same chair, same jumper, wall gone. Total spend: nothing. It doesn’t look flawless — I currently have the lighting of a man who’s been composited into his own house — and I’m not fixing the old ones retroactively. This is a channel about learning in public, which means the early stuff stays visibly early.

None of it has made a dollar.

Not a cent. Twenty-four people on this list, one of whom is me. Ten subscribers on YouTube. Three hundred and twenty-nine views across everything I’ve put on there. Money going out every month, and nothing whatsoever coming back.

And on Thursday I found out that one of the things on that list had never worked. The system that turns a comment on Instagram into a subscriber here — live for four weeks, collected two email addresses, both of them mine. One missing word, in one setting, in one step I never looked at. No error. No warning. Nothing red anywhere. The dashboard said live.

Green doesn’t mean working. Green means nothing is reporting a failure. Those are not the same thing and I had never once thought about the difference.

So the list above isn’t a victory lap. It’s an inventory, and some of it was quietly broken while I stood there admiring it.

The thing I’d tell someone standing on day one.

Here’s what actually surprised me, and it isn’t “AI is amazing”.

I didn’t research any of this. I didn’t compare platforms, or read reviews, or find out what people in my position are supposed to use. I described the problem out loud, in ordinary words, the way you’d explain it to someone in a pub — and something else picked the tool. ManyChat and Buffer both arrived that way. Left to myself I’d have spent a fortnight choosing and still got it wrong.

That’s the actual leverage, and it’s not the bit anyone advertises. It isn’t that the machine writes things. It’s that not knowing what a thing is called has stopped being a reason you can’t use it.

The judgement is still mine, and it has to be — half of last week went on finding out where I’d been confidently wrong, and no software found those for me. But the distance between “I’ve had an idea” and “there’s a thing on the internet” has gone from months to an afternoon. I’m fifty-five. That is the single most interesting fact I’ve learned this year.

The real numbers this week:

A correction to my own numbers.

I’ve been publishing my monthly burn as a hundred and seventy-six dollars. It’s wrong. Not enormously, but wrong, and I’d rather correct it here than let it quietly fix itself.

The tool that generates the ten-month-old isn’t in that figure. I bought it months ago for my other business — Tidy Design, an Australian streetwear store I run — so when I listed what The Long Game costs to run, it simply never occurred to me. But it’s making Baby Mark now, and a cost you’re using is a cost you’re carrying. So from this month half of it sits here: fourteen dollars fifty a month, taking the real number to a hundred and ninety dollars fifty.

Nobody would ever have noticed. That’s rather the point.

And about the links that pay me.

Some of the links above earn me a commission. So far they have earned me nothing at all — the running total is nought — but they’re worth explaining, because they aren’t a sideline bolted onto this. They’re the first honest attempt at the actual point of the whole experiment.

The question I’m trying to answer is whether someone starting from zero, in his mid-fifties, can build real income with AI doing the heavy lifting. Affiliate commissions are the cheapest possible test of that: I use a tool, I tell you what it did and what it cost, and if you go and use it too, I get a cut. There’s no course, no coaching, and nothing invented to sell you. If a tool stops earning its place I’ll stop linking to it, and you’ll read about that here as well.

Four programmes have said yes. One has said nothing since the 18th of August. Netlify said no and told me to come back at a thousand visitors a month, which is a fair answer and a checkable one — I’ll reapply when the number says so rather than when I feel like it.

Instagram went from 47 to 120 in three days. Facebook went from 16 to 32. I can’t tell you what caused either of them, and last week I wrote an entire paragraph about Instagram going down by one. That’s twice in this email I’ve had to admit I’m reading numbers I don’t understand yet. I’d rather tell you that than dress up a guess.

The one I keep going back to is video one. It went up ten days ago. Fifty-two views on Tuesday, eighty-one on Thursday, a hundred and forty-three yesterday, a hundred and sixty this morning. I haven’t promoted it, paid for it or asked a single person to watch it. Something out there has decided to keep handing it to people.

It is also, by a distance, the worst thing I’ve made. It takes thirty-three seconds to get round to telling you what the channel even is. The two I’ve made since are better in every way I can measure and they’re sitting on thirty-seven and thirty-one.

And then there are the Shorts. Three of those six things on YouTube are Shorts — vertical, under a minute, cut from footage I already had. Fifteen views, eighteen views, and sixty-eight. That last one has been watched by more people than either of the videos I spent hours editing.

Put it next to ninety-eight TikTok followers from three posts and I have two completely separate signals pointing the same way, which is more than I’ve had about anything else this month. The short stuff travels. The long stuff is where nearly all of the work has gone.

I’m not rebuilding the plan on a sample this small. But I’d be an idiot not to notice.

One more thing, and it’s the one that actually got me. Last week I had a tidy explanation for video two doing badly: I’d forgotten to tick a category box when I uploaded it. Two days on, video two is at thirty-seven and video three, which went up after it, is at thirty-one. It didn’t just catch up — it went past. Whatever that dropdown did, it wasn’t that.

That’s the fourth time this month I’ve explained something confidently and turned out to be wrong. I’m starting to think the skill I’m actually learning here isn’t AI, or video, or any of the tools. It’s waiting a bit longer before deciding what a number means.

And the honest one: I still can’t point at a single subscriber and tell you a video sent them.

But as of yesterday I can point at one and tell you exactly where she came from.

Someone I have never met commented on a post, was sent a message, typed in an email address, and landed on this list. Six minutes, start to finish, nobody involved at any point.

That’s the funnel from the top of this email. The one that spent four weeks doing nothing at all. I repaired it on Thursday with no evidence whatsoever that anybody would ever use it, and on Saturday somebody did.

Attention still isn’t my constraint. But for the first time in thirty days I know where one person came from, which is one more than I could tell you about the other twenty-three.

If you’ve not met him yet.

Video three is the introduction to the ten-month-old, and the best line in it is his: “You didn’t build me. You subscribed.” He’s allowed to be wrong, which matters more than it sounds — the moment he’s always right, this becomes an advert for AI instead of an honest account of using it.

Video two is here — a landing page built in under four minutes. It goes wrong twice. Both times are still in it.

One thing before you go.

Hit reply and tell me: what’s the one thing you’d hand over to an AI tomorrow, if you trusted it enough?

I’ll go first. Scheduling social media. I would rather do almost anything else, and for thirty days I haven’t had to.

I read every reply.

— Mark

The Weekly Escape Report documents one person’s public experiment to build financial and location freedom — every number, every failure, every move. Free, every Sunday.

→ longgameco.com/newsletter

This newsletter runs on beehiiv. That’s an affiliate link — if you start a newsletter through it, I earn a commission. I’d have recommended it anyway; it’s what I chose before there was any commission in it. I’ll always tell you when a link pays me.