AI coding tools have become part of everyday development.
Cursor, Claude Code, Gemini CLI, Codex, and many others each have strengths.
The challenge isn’t capability.
The challenge is continuity.
The Context Fragmentation Problem
A typical workflow might look like:
- Generate code in Cursor
- Review architecture in Claude
- Experiment in Gemini
- Return to VS Code
The codebase remains consistent.
The project remains consistent.
The context does not.
Each tool starts with partial knowledge of previous decisions.
Developers repeatedly reconstruct information that already exists somewhere in the workflow.
A Different Perspective
Most AI tooling focuses on improving the assistant.
We’re exploring another layer:
What if context itself became portable?
Instead of asking:
“Which AI should I use?”
We ask:
“How can I switch between AIs without losing project understanding?”
What We’re Experimenting With
Contorium is an open-source attempt to make project context persist independently of any specific AI tool.
The goal is not replacing Cursor, Claude Code, Gemini, or future assistants.
The goal is making them interoperable.
A project should outlive any individual model.
A workflow should survive ecosystem changes.
Open Question
As AI development tooling evolves, do you think the future belongs to:
- One dominant AI platform?
- Multiple specialized assistants?
- A shared context layer connecting them all?
I’d love to hear how other developers think about this problem.
Website: https://www.contorium.dev/
https://github.com/ContoriumLabs/contorium
Top comments (0)