Why we cancelled our entire SaaS stack, and how we built what replaced it

A few months ago, we did something that felt reckless: we cancelled our CRM, our email automation, our website builder, our prospecting tools. All of it.

Right now we're a remote team of seven. We've built companies before, and we know how startups are supposed to work: You buy tools, you learn them, and you operate them. That's been the game for over a decade.

But, as many in the terminally-online startup community are acknowleding, something big has shifted.

Why the best AI users are 10–100x better than everyone else

We'd been working on a different product deep in voice AI infrastructure. Months of exploring how people interact with AI systems through conversation.

What we kept seeing: the top 1% of AI users get incredible results. They've mastered prompting, context-setting, building the right workflows. They're 10x more productive. But most people don't have the time or technical skill to get there. They're running businesses, not tinkering with AI setups.

The gap isn't intelligence. It's context. To be genuinely useful, AI systems need to understand your business: your goals, customers, challenges, how you work. The best users figured out how to give AI that context. Everyone else is stuck on basic ChatGPT.

Voice, we realized, is a natural way to give AI systems rich context. You can explain your business in a conversation far more easily than filling out forms or writing prompts.

As we started talking to more founders, operators, and leaders who were trying to use more AI in their work, the pattern became clear. They weren't struggling with AI's capabilities—they were struggling to give it the right context at the right moment. Finding less technical ways to solve that problem felt like one of the highest-leverage things we could work on.

The moment we cancelled everything

Andrej Karpathy posted this in late 2025:

14 million views. It resonated because developers are living it.

I experienced it firsthand. I needed a CRM tailored to how we actually sell. Not the bloated thing we were paying for (which was also full of features we'd never touch). I knew exactly what I wanted our CRM to do for us, and it was a far, far simpler piece of software than what we were paying for.

So, I opened Claude Code and got to work. By the end of the day, I'd built us a new CRM that had all the features we needed, and nothing else. We called it Toyo.

This wasn't a prototype. It was production software, built exactly for us.

The cost to create software has collapsed. For new projects, it's now harder to learn an existing tool than to describe what you need and have it built.

So we kept going. Website builder and CMS? Cancelled. My co-founder rebuilt our marketing site in a weekend (and it was better than the old one). Prospecting tools? Rolled into Retro. We built exactly what we needed, and nothing more.

Within weeks, we'd replaced most of our stack with custom tools tailored to how we work.

The question: Why were we still doing the work?

Then one morning, I was using the CRM I'd built to research prospects and enrich some leads.

And a question hit me: why am I doing this?

Claude Code built the system in an afternoon. It understood databases, APIs, business logic. If it could build the CRM, why couldn't it operate the CRM?

I gave it direct database access, a browser, and tools. Instead of me using the interface, it queries the database, researches companies, and writes updates directly. It doesn't tire. I can run multiple instances in parallel.

Within weeks, we threw away the interface entirely. The team uses a simple database viewer to check the pipeline. Agents do the operational work. We review and direct. We don't operate.

What happens when agents use software

Something clicked for me: SaaS isn't dying because of AI. It's dying because of what AI makes possible.

The old model assumed humans would always be the operators. Vendors build tools. You buy licenses. You adapt to how someone else thinks you should work. Interfaces exist because humans are slow and need visual layouts to understand data.

But what happens when the operator isn't human? Aaron Levie put it perfectly:

Software was valuable because it helped humans work. Agents are valuable because they do the work. If agents do the work, the interface layer suddenly looks a lot like overhead.

"AI kills SaaS" is the headline right now, but it misses the point: Software is about to become more abundant than ever. What's dying is the SaaS business model that sells one product to millions of people. What's emerging is software that fits each business, built on demand by agents, operated by agents that get real work done.

Our experience becoming an AI-native team

Today, our small team can operate like a much larger company: Teammates who'd never written code before are now building their own tools to automate their own workflows. Campaigns that would have taken days can now launch in hours.

We know this works because we're living it every day.

But here's what became clear: the opportunity to help less technical founders and teams get maximum leverage from AI is far larger (and far more meaningful) than what we'd been working on. And our team, with decades of experience building tools for operators, was well-suited to build it.

What we're building

Every team we talk to is stuck on the same things.

• AI tools are too technical for most of their people.
• Their data is scattered across dozens of apps, so AI can't help even when they want it to.
• The automation they actually need still feels out of reach.
• The overall pace of change in AI tools is so relentless that nobody has the space to figure any of it out.

So we shifted. We sprinted. The problem was clear: the best AI users are 10-100x more productive because they've mastered giving AI the right context. Everyone else is stuck. We decided to build the tool that closes that gap and help any founder or operator get those results without needing technical skills.

That's Toyo. An AI computer in the cloud where you describe what you need and agents do the rest. No engineers required.

We're building everything we learned about where voice AI works today into the core to how Toyo works. The easiest way to give Toyo rich context about your business is to just talk to it. Explain your goals, your challenges, how you work. The agent remembers, gets smarter over time, and starts delivering the kind of leverage that today only the top 1% of AI users get.

We're now onboarding our first users. If you're a founder or operator ready to stop operating software, we want to talk: Get early access to Toyo.