Translation is easier than ever and if you are in a more modern company where the barriers to adapting the tech stack (even things like deploying Claude code at scale) I agree the value is minimal because you can build something that does this. But in large established enterprises who are still transforming and often have multiple BI tools - a standard that improves interoperability removes one friction point. Right now so many tools don’t even expose their semantic layer well in any format - sometimes you can only build it through clicks in a UI. Which makes it hard to adopt because you can’t build a semantic layer all in one tool, but it’s tricky to make them talk to each other and the warehouse. So moving towards all code based reduces friction. And the closer the formats are, even better.
Yes this is why I still think it’s valuable. Even though it’s easy for me to go between Hex, dbt, Databricks and Thoughtspot right now… it seems unnecessary. At the same time even if they did all support the same spec we would still have to integration work.
Funny thing is, the true value for AI is in semantic modeling, the ontological type. Sure, text literal metadata is great, but lacks context and meaning, which is the real value prop from actual semantics.
I had the same reaction. What is the value of a standard YAML representation in a world where AI can already understand any kind of textual representation of the same metadata? It feels to me like all these semantic layers are still pursuing the goals of a pre-AI era.
Honestly, I was a big fan of the idea of universal semantic layers and standardization efforts pre-AI. If AI will be doing a lot of the heavylifting in helping me analyze data, why would I need my dashboarding solutions from the 2010s to be able to talk to my universal semantic layer? Even my excel?
Translation is easier than ever and if you are in a more modern company where the barriers to adapting the tech stack (even things like deploying Claude code at scale) I agree the value is minimal because you can build something that does this. But in large established enterprises who are still transforming and often have multiple BI tools - a standard that improves interoperability removes one friction point. Right now so many tools don’t even expose their semantic layer well in any format - sometimes you can only build it through clicks in a UI. Which makes it hard to adopt because you can’t build a semantic layer all in one tool, but it’s tricky to make them talk to each other and the warehouse. So moving towards all code based reduces friction. And the closer the formats are, even better.
Yes this is why I still think it’s valuable. Even though it’s easy for me to go between Hex, dbt, Databricks and Thoughtspot right now… it seems unnecessary. At the same time even if they did all support the same spec we would still have to integration work.
Isn’t this simply a YAML data contract?
Funny thing is, the true value for AI is in semantic modeling, the ontological type. Sure, text literal metadata is great, but lacks context and meaning, which is the real value prop from actual semantics.
Move it beyond analytics already!
I had the same reaction. What is the value of a standard YAML representation in a world where AI can already understand any kind of textual representation of the same metadata? It feels to me like all these semantic layers are still pursuing the goals of a pre-AI era.
Honestly, I was a big fan of the idea of universal semantic layers and standardization efforts pre-AI. If AI will be doing a lot of the heavylifting in helping me analyze data, why would I need my dashboarding solutions from the 2010s to be able to talk to my universal semantic layer? Even my excel?