For years, J.C.P Laboratory was known as a cloud infrastructure consultancy. We managed servers, built Kubernetes clusters, handled migrations, and kept businesses running. We were good at it.
And that was the problem.
The trap of competence
When you're good at something, people pay you to keep doing it. The revenue is steady. The work is predictable. But predictable work has a gravitational pull — it slowly consumes all the space that should be reserved for the unpredictable, the research, the things that don't have a clear business case yet.
The lab was drifting toward being a consultancy with a research hobby. That's backwards.
The split
We moved all managed services to Pancake Solutions — a separate company with its own team, its own brand, and its own focus. J.C.P Laboratory is now a pure research lab: AI, language, and open-source tooling.
Pancake handles the infrastructure work that used to live here. The lab gets to focus on what it was supposed to do from the start — build things that don't exist yet.
What we learned
- Brand matters. When your website lists "managed Kubernetes" next to "AI research," nobody believes either one.
- Revenue is not identity. Just because you can make money doing something doesn't mean you should build your identity around it.
- Saying no is a competitive advantage. The hardest thing we did wasn't technical — it was turning down work we were perfectly capable of doing.
The result
The lab is smaller, more focused, and more productive than it's been in years. Astra Agent exists because of this decision. The Kreole LLM project exists because of this decision. Sometimes the best way to move forward is to let go of something you're good at.