Marivo / Blog
The design, practice,
and evolution of Marivo.
Essays on product design, field experience, release changes, and the community around Marivo. We explain the decisions behind the system and examine how they hold up in real data projects.
Evidence Engine: Results Are Not Conclusions
A result can be correct and still be overstated in the retelling. This article explains how Marivo makes evidence traceable and keeps every claim within what the result actually supports.
Read article
Archive
More articles
Analysis DSL: Constrain Analytical Actions, Not Agent Reasoning
How Marivo makes semantic objects govern each analytical action while leaving hypotheses, direction, and interpretation to the agent.
Read article
The Semantic Layer: Turning Business Meaning into an Executable Contract for Agents
Why agents cannot derive approved business meaning from schemas alone—and how Marivo uses semantic objects, Python, Git, runtime validation, and CI to keep that meaning dependable.
Read article
Why Data Analysis Agents Need a Harness, Not Just SQL Generation
How governed semantics, typed analysis, persistent sessions, and traceable evidence give open-ended agent analysis dependable boundaries.
Read article