From Side Project to AI Skills Business: An Honest Retrospective

I have been building RemoteOpenClaw, an open marketplace for AI skills and personas, and I want to share what the journey has actually looked like — the wins, the mistakes, and the things I wish I had known earlier.

How It Started

I kept building the same AI capabilities for different clients. PDF extraction, report generation, domain-specific personas — the same patterns, different contexts. One day I asked: why am I not packaging these as reusable components?

That question led to RemoteOpenClaw.

What Went Right

Building on an Open Standard

Choosing to build on the OpenClaw standard rather than a proprietary format was the best early decision. It meant:

  • Skills are portable Markdown files, not proprietary blobs
  • Users feel safe investing time because there is no lock-in
  • The community contributes because the format is accessible

The 90/10 Revenue Split

Most marketplaces take 30%. We went with 90/10 (creators keep 90%). The math is simple: AI skills are lightweight digital goods. Our marginal cost per transaction is near zero, so taking a huge cut would just be extractive.

This decision attracted early creators who had been burned by platform fees elsewhere.

Starting with Quality Over Quantity

We could have launched with hundreds of low-quality skills scraped from various sources. Instead, we started with a smaller catalog of vetted, well-documented skills. This built trust early.

What Went Wrong

Underestimating the Chicken-and-Egg Problem

Every marketplace faces this: buyers want selection, sellers want buyers. I spent three months building features before realizing I should have been building community.

Over-Engineering the Platform

The first version had skill versioning, dependency resolution, automated testing, and a dozen other features nobody asked for. The second version stripped most of that away. Users wanted: browse, buy, download. That is it.

Pricing Confusion

I originally let creators set any price. The result was chaos — similar skills priced at $5 and $500 with no clear differentiation. Adding pricing guidelines and category benchmarks helped.

What I Would Do Differently

  1. Launch with 10 high-quality skills, not a platform — the marketplace infrastructure can come later
  2. Build in public from day one — I waited too long to share progress
  3. Talk to 50 potential users before writing code — I assumed I knew what people wanted
  4. Focus on one vertical first — trying to be everything for everyone slows growth

Where Things Stand Now

The marketplace is live at remoteopenclaw.com. We have a growing catalog of skills across content, data, development, business, and support categories. The creator community is small but engaged.

It is not a success story yet. It is a work-in-progress being shared honestly.

Advice for Other Builders

If you are thinking about building a marketplace or platform:

  • Solve your own problem first — be your own first user
  • Choose open over closed — it attracts better contributors
  • Launch embarrassingly early — you will learn more from real users in one week than from planning for one month
  • Measure what matters — active creators and repeat buyers, not vanity metrics

Building something similar? I would love to compare notes. Drop a comment or reach out.