Hello {{First name|Predictable Revenue community}},

Predictable update: onboarded 3 more users this week and the feedback is flying in so quickly I can hardly keep up. It’s really fun to get the gift of “here’s why I hate this part of your product”. Emotions are good, it means they care enough to be mad, which I think is a strong sign of being onto something.

I got on a call recently with a client who's one of the smartest people I get to talk to. When it comes to intellectual horsepower, I was outmatched. I usually am… but especially with him.

He's worried that software as a business model is going away. I don't share that fear. But I couldn't fully answer his argument on the call, so I'm writing this to sharpen my own thinking. You get to watch.

It started a few days earlier, on my bike.

I was halfway up a hill when I realized I could steal Clay's customers. Not all of them. But a lot of them. And it would take two features and one Claude Code skill.

The skill logs into a customer's Clay account, reads the play they've built, and rebuilds it inside Predictable. The customer doesn't have to do anything except say yes.

A year ago, that was near impossible (I tried). A Clay workspace someone spent months building is exactly the kind of thing that keeps you from leaving. It's architecture. You don't rip out architecture. Now it costs basically nothing to move.

It's not just Clay. Moving from GCP to AWS used to take 6 to 12 months and a dedicated team. Now it can be done in a few days.

I've done it to my own CRM. I built Yaletown, a fork of Gas City, which is a derivation of Steve Yegge's Gastown. I plugged in my email, my calendar, my call recordings, and most importantly Jev, and it sorts everyone I talk to into Meet, Disco, Manage, and Nurture. Every CRM vendor I've ever paid was counting on me never doing that.

In Hamilton Helmer's 7 Powers, switching costs are one of the seven sources of durable advantage. That one is eroding fast.

So I brought my answer to the call. If switching costs are going away, the moat must be expertise.

My bet with Predictable has been that the value isn't the code. It's the judgment. There are thousands of decisions and hundreds of hours of thinking baked into how it works. Someone could write better code than me. Most people probably do. But I'd build better plays, because I've spent 14 years in this space.

He didn't buy it.

His core point: learning is cracked. If an LLM can be trained on your domain, your knowledge of the space stops being the advantage.

Once you put your expertise into software, he said, it's cheap to rip off. The copier doesn't need your creativity. They just need your output.

Then he told me about a friend's company. Ex-DeepMind folks that built their own LLM specialized in CAD producing models. Then Astra learned CAD, and their positioning collapsed. They spent years of their life building specific expertise only to have its significance disappear overnight. That's process power, another of Helmer's seven, dissolving because learning is cracked.

Take that far enough and you land where he did. If switching costs are gone and a model can learn your domain, maybe there's no durable software business left to build.

Some of what he said, I agree with. Things are accelerating. And go-to-market isn't a source of differentiation. I learned that one the hard way. When we started doing cold email, we figured it out before most people did, and customers rained out of the sky. Then everyone figured it out and our top of funnel slowed right down. GTM has always been a cold war, and every edge erodes. What's new is how fast.

Here's where I part ways with him.

He's right that some of Helmer's powers are eroding. Switching costs are going. And LLMs are speeding up the erosion of process power and cornered resources, at least when the resource is hyper-specific expertise. That's what happened to the CAD team.

But the rest still hold. Scale economies and network effects aren't going anywhere. Brand might matter more than ever, because when anyone can build a DIY version of anything, a name people trust gets more valuable. And at least one of the powers might be easier to build now than it's ever been.

Features aren't power. I'm building things in Predictable I never thought possible, and I've accepted that none of them are a moat on their own.

Speed is different. It's not power, but it's an advantage, and right now a big one. As a startup competing against an incumbent, your real advantage is reaction speed. You can ship state-of-the-art faster than they can.

Ever tried to swat a fruit fly out of the air? It's not impossible. But it's not easy.

That's every startup. Incumbents are bigger. We're faster and more annoying. But you have to know you're the fruit fly and not pretend you're the one doing the swatting.

And speed only matters when things are changing. When there aren't many kids in the innovation ball pit, nothing moves and speed doesn't buy you much. When it's packed, like right now, speed gets you from zero to one.

But speed isn't power. Helmer defines power as the ability to generate persistent differential returns. Speed can get you differential returns. It can't make them persist.

And yes, if learning is cracked, my 14 years are worth less than they used to be. But they're still a head start. Your experience is the base you build that first edge on. It gets you a little ahead of all the other startups that are also moving fast. If you're lucky, it compiles into something the market cares about. But if you don't turn that lead into one of the real powers, you won't last long. Fruit flies don't.

Maybe that's the fruit fly in me talking, and I'm wrong. Or maybe I just have a more optimistic worldview. I think some companies built on a one-time edge are going away. I don't think software is.

So I'd argue 7 Powers has never been more relevant. If switching costs won't protect you, you have to be deliberate about the powers that still will.

For startups, Helmer points to counter-positioning: a business model your competitors can't copy without losing money. That's the one I think AI makes easier. It's changing what buyers expect and what's possible to sell, and that opens up business models that didn't exist a couple of years ago.

Here's what that looks like for me right now. Yaletown pulls the action items out of my call transcripts, emails, and calendar. It doesn't replace a CRM. It's the reason people use a CRM in the first place, especially an AI CRM.

I haven't decided how to take it to market. I could package it as SaaS, like everyone else. Not differentiated.

I could sell it like boxed software. Version 1 is $99. I update it every year, and you buy the new version. You bring your own tokens through your Claude Max or Codex subscription.

Or I could make it work really well with Predictable and open source it, so more people end up on Predictable.

Those last two are not things an AI CRM like Clarify would be excited to see in the market, and I don't think they could profitably copy them. I'm speaking at Clarify's Horizons conference on December 9th in Seattle, for founders and GTM leaders who want to know what's next for GTM. Assuming Patrick doesn't uninvite me after reading this.

So here's my question for you. How could you structure your business model, your pricing, in a way your competitors couldn't copy without losing money, but that would still be profitable for you?

Collin

PS - thanks for still reading. Since you’re still here, here’s something fun for you, go buy an ESP32-S3 development kit (with a screen). They’re $20 and you can program them super quickly with Claude Code.

The Terrifying Art of Finding Customers is the playbook for founders who never planned to do sales but have to figure it out anyway. It pulls together 13+ years of building software startups, building outbound systems for hundreds of B2B startups, and more failed experiments than I'd care to admit, into a practical guide for finding your first 10, 100, and 1,000 customers.

It will help you build a repeatable way to find customers without losing your weekends (or your soul).

Order now and you'll unlock a stack of exclusive bonuses I actually use with my consulting clients: Email me the receipt and I’ll share the resources.

Email me your receipt and I’ll share the resources.