We built the wrong product

Spinup v1 was an OpenClaw hosting service. The pitch was simple: click a button, get your own agent. No terminal. No config files. No SSH. OpenClaw for normies.

We were wrong about almost everything.

"And now what?"

OpenClaw was everywhere. Everyone was talking about it. So when we opened the beta, people signed up fast. They wanted to see what the hype was about.

The problem showed up within days. People would spin up an agent, poke around for a bit, and then nothing. The most common question we got was some variation of "cool, and now what?"

That question kept coming back.

Because here's the thing. Out of the box, a harness like OpenClaw doesn't do much on its own. Its power comes from giving it systems, skills, CLIs, code, and access to things. That takes configuration. That takes knowing what you want it to do.

Most of our beta users signed up because of the hype, not because they had a problem to solve.

They wanted magic. We gave them a blank canvas.

The support math

We had 25 beta customers. I was spending about an hour a day on support. Daily WhatsApp messages. "Can my agent do X?" "How do I access Y?"

The questions made sense individually. But the volume didn't. An hour a day for 25 users is not a ratio that improves with scale.

The architecture made things worse. We were spinning up a dedicated VPS for every single agent. Great for isolation. Terrible for everything else.

When we realized our base image was missing critical packages, we had no centralized way to roll out an update. Our security was properly hardened, which meant we couldn't just SSH into each server. Good from a security perspective. Bad when you actually need to fix something.

The process: put the VPS in rescue mode. Grab a temporary root password. Log in through our VPS provider's clunky web CLI. Run the update manually. Per server. One by one.

That was the moment we looked at each other and thought: this does not work.

The competition we couldn't outspend

While we were debugging rescue mode workflows, well-funded competitors were hosting meetups with 400+ attendees. Serving lobster and crab. Running ad campaigns we couldn't touch.

Everyone wanted to jump on the OpenClaw hosting bandwagon. Big hosting companies with huge existing customer bases. Startups with deep pockets. Indie hackers who got there before us. The space filled up fast.

Two indie hackers from Amsterdam weren't going to win that spending war. We had no first-mover advantage and no financial runway to buy one.

We needed a different game.

The audience mismatch

The deeper problem was who we were building for.

"OpenClaw for normies" sounds great in a pitch deck. No setup hassle, no configuration hell. Just click and go.

In practice, there's a floor of complexity you can't design away.

Take connecting an agent to Discord. We thought our interface made it trivial. But the user still has to go to Discord, create a bot application, navigate permissions, copy tokens, paste API keys. We've done this dozens of times. For someone who hasn't, it's overwhelming.

Tokens. API keys. Access controls. Permissions. OAuth scopes. We kept simplifying the UI. The underlying complexity didn't budge.

There was a constant mismatch between our technical brains and the non-technical audience we were trying to serve. It showed up in every onboarding flow, every support conversation, every feature we shipped.

What changed

The pivot came from a simple realization: we should build for people like us.

Spinup is now a runtime for developers. One agent. One computer. Any harness. The Spinup Agent is the durable object: files, packages, and state stay with it. Swap the harness without rebuilding the setup. Instead of a dashboard where normies click buttons, we have an SDK, a CLI, and an API. Instead of hiding complexity behind a consumer UI, we're building tools for people who already know what they want their agents to do.

Running spinup agents list in the terminal feels right in a way the old product never did.

We can dogfood our own product now. We write docs in a language we actually speak. When a user asks a question, we share enough context to understand it immediately.

Building for builders. Closer to who we are.

What we learned

Four things, mostly obvious in hindsight:

• Hype is not demand. People signing up because something is trending is not the same as people signing up because they have a problem to solve.
• Support load at small scale is a signal. If 25 users cost you an hour a day, the product has a friction problem. Not the users.
• Don't fight spending wars you can't win. Find a different angle.
• Build for an audience you understand. The language gap between you and your users shows up everywhere. Every onboarding flow, every support thread, every design decision.

We got it wrong. Then we fixed it. That's the whole story.

---

Spinup is pre-launch and taking shape as a cloud agent runtime. If you want to follow along as I build, find me on X: Maurice.