I sound the same as everyone else now. So do you.

I reread MyPip's homepage last week and felt a bit sick.

Not because it was bad. Because it was fine. "AI-powered lead generation for founders." Clean, competent, forgettable. I could paste any of forty other logos above it and not one person would notice the swap.

I wrote back in December that this positioning was wrong. I thought the problem was the words, product category versus workflow, all of that. It wasn't the words. It was that a machine could have written them, and if I'm honest, one basically did.

Think about what an AI actually does when you ask it to write your marketing. It's read everything ever written, and it hands you back the most probable version of what you asked for. The most probable version of anything is the average of it. That isn't a bug you can prompt your way out of. That's the whole design.

Now do that with paint. A hundred founders, each holding their own colour. One's a deep red, one's a bright yellow, one's a strange green nobody's mixed before. Every one of them tips their tin into the same bucket, because the bucket is free and fast and genuinely very good. What comes out?

Brown. It's always brown.

Average is brown.

For half of what a founder does, the bucket is a gift. I couldn't really code three years ago. Now I've shipped a product that took two years and most of which I could never have written by hand. AI took a below-average developer and dragged him up to average, because writing software is consistent and learnable and there isn't much nuance in a for-loop. Up was exactly the direction I needed to go.

But marketing isn't a for-loop. Good marketing stands out by definition, and the one thing an average can never do is stand out. So the same machine that lifts your code to the middle drags your words to the middle, and the middle is the one place nobody remembers you were.

Remember old MailChimp? That loud, aggressive yellow. The winking monkey. You knew it from across a room. A few years back they rebranded into tasteful, rounded, sensible SaaS, and now I couldn't pick their landing page out of a lineup of thirty. They did that to themselves, on purpose, with a design budget. AI does the same thing to all of us at once, for nothing, while we thank it for saving us the afternoon.

Everyone told me AI would flatten the playing field. Solo founder gets the same tools as the funded team, the gap closes. It did the opposite. I watched it happen through the whole vibe-coding wave. The get-rich-quick crowd turned up, found everyone standing on the same flat ground holding the same tools, realised there was no cheap edge left to grab, and wandered off. The ones who quit quit because grinding is hard, and no tool has ever fixed that. And the founders with the thing you can't download shipped more and pulled further ahead than they'd ever been.

The tools didn't narrow the divide. They showed you who was already on which side of it.

Because building stopped being the proof of anything. What used to take months takes a weekend. New business registrations climbing, new repos climbing, apps pouring out. And the median revenue across all of it sits dead flat. Everyone can build. Almost nobody can get a single person to care. The AI coding companies sold you the dream of building anything by dinner, and the moment you believed them you inherited their real problem, which is that so can everyone else.

So what actually survives the bucket?

Only the colour that was never in it. Your specific experience, the thing that happened to you that hasn't been written down ten million times for a model to average out. The beta user who broke my product by replying 450 times in four days and taught me to send fewer, better messages. The Plastic Bank customer profile we were completely certain about and got completely wrong. That's the green nobody's mixed. A machine can tidy my grammar, but it was not in the room, and being in the room is the entire asset now.

That's why I've stopped holding things back. A half-formed idea posted in public gets air, gets someone telling me they disagree, and the disagreement is what finds the angle. The same idea run through the bucket first shows up grey and already dead. Marc Lou posts his revenue every month and runs nine little apps off the back of it, and it works because people trust him. The apps are almost incidental. You can't average a person.

This is the part I can't square.

I build an AI product. Its entire job is to research people and help you say something to them. And here I am telling you AI drags everything to the mean. So what stops MyPip averaging its own users into the exact same grey outreach clogging every inbox already? Get that wrong and I'm just another tin emptying into the bucket, with a tidier landing page than most.

I have a bet. The bucket only averages what's already public and generic. Aim it instead at the buried, specific, slightly weird detail of one real person, the thing their job title will never tell you, and you're not pulling toward the mean anymore, you're pulling toward them. Research as the un-averaging move. That's the whole wager the product stands on.

I think it's right. I've watched it work in my own conversations. But I felt just as sure about that homepage, and a machine could have written that.

So I'm rewriting it this week. This time it needs to sound like it came from someone who was actually there.

More soon.