From question to experiment
An experiment starts with a hypothesis and an evaluation criterion. It may explore a simpler integration, a document assistant or a workflow that removes a manual task.
Areas of exploration
- RAG and retrieval: verifiable answers, source selection and access control.
- Agents and tools: small operations, explicit permissions and failure evaluation.
- Automation: exception handling and human control where needed.
These are research themes, rather than announced products or public demos.
Available reading
The private RAG architecture and MCP versus code execution provide context for these themes.