We built the first AI tool podcasters ever had. Then we killed it ourselves.

A few months later, ChatGPT launched - and the wave of AI tools and competitors it opened the door to started chipping away at what made Capsho different. What used to require a dedicated product was now something anyone could get, in some form, for free, from the AI itself.

We spent the next few years trying to out-build that. More features. More scope. A bigger, “smarter” Capsho. It worked for a while. But it kept getting harder, and the gap kept closing.

The first was the category itself. Building a product inside generative AI, competing on how smart the AI feels, is a race against models that get better every few months, from companies with far more resources than we had. We couldn't out-build the frontier. Nobody building a wrapper can, forever.

The second was harder to admit: the SaaS model itself was the problem. To make Capsho scale, we'd built a fixed, deterministic workflow - and every customer had to learn our way of doing things, whether or not it fit how they actually worked. We weren't just competing with the AI underneath us.

We were doing to our own customers the exact thing we now warn people about: renting them a process, instead of giving them one that was theirs.

Back then, building anything meant paying an outsourced dev team and waiting on them. When Ash started directing AI to build instead - no developers to hire, no queue to wait in - that changed everything.

That shift is what let us build everything else, too - client platforms and sites, and our own ventures: Tuulie, Perkeroo, MixerooCrew, and others. All of it built the same way: with Monty and his team, not a SaaS dashboard.

It was going well. It was also, honestly, fun - in a way the SaaS grind hadn't been for a long time. We weren't forcing anyone into a workflow. We were building the actual thing each project needed, with a team that already knew how to do the work.

Deirdre - Capsho's co-founder - started as non-technical as it's possible to be. Today, she runs Do The Thing's entire build stack this way: automations, websites, client hubs, affiliate hubs, all built and run through the same team that's now Monty.

Nobody taught her to code. She didn't need to.

She needed a team that could - and the judgment to direct it.

We sunset the SaaS. Not because it failed - it did real work for real people, for years. We sunset it because the model underneath it couldn't be fixed with more features, and because we'd already found something that worked better, sitting right next to it the whole time.

We also didn't do right by the people who trusted us with the first version. When we wound the SaaS down, we let it go quiet instead of giving our customers a real heads-up or a real next step.

That's on us.

So this time we're doing it in person, not in an announcement: going back to the people who trusted us with the first version, one at a time.

Capsho, as it exists now, isn't a rebuild of the old product. It's what we'd already been doing for clients and for ourselves, made into something anyone can own.