Data Engineering Weekly

The episode delivers a handful of genuinely useful distinctions - the compiler-as-differentiator between knowledge graph and semantic layer, AI as a maintenance mechanism, and the idea that organisational scale not data scale is the real trigger - but a large portion of runtime is spent on foundational explanations and conversational meandering rather than novel claims per minute.

“every semantic layer is a knowledge graph, but not every knowledge Graph is a semantic layer. And why is that? It's because a semantic layer uh, in my mind always has a compiler.”

“you can't have metrics without a semantic layer, right?”

There are a few genuinely fresh framings - the compiler as the definitional dividing line, the forcing-function argument about AI access pressuring data hygiene, and the claim that MCP dissolves semantic layer standardisation wars - but the bulk of the episode recycles well-known BI lock-in and single-source-of-truth arguments that have circulated for years.

“I think that the semantic layer could become invisible right in the future”

“if you're doing MCP and it doesn't have to be mcp, it could be some other agent standard, I don't really care. Let's say MCP for now. If you're doing mcp, the format of the semantic layer and the standard of the semantic layer doesn't matter.”

David is a genuine practitioner with 15+ years of hands-on experience, having led data teams at Worldpay, co-founded Avora and Delphi Labs (a semantic-layer AI company), and worked inside Cube and Metaplane; he speaks from real implementation experience rather than thought-leadership positioning, though he is not a widely recognised industry figure at the top tier.

“I ended up leading data teams at various companies including List, Elevate, Credit, worldpay before going on into startups where I've worked, been a co founder a couple of times and also worked at companies like Metaplane and more recently Cube and have founded a company called Delphi Labs in the past which was focused on applying AI to semantic layers.”

The episode names specific tools and vendors (Cube, Looker, Malloy, DBT, Databricks metric views, DAX, MCP) and offers a couple of illustrative examples like the ACV formula and a 30-second vs 30-minute latency comparison, but there are no hard metrics, customer case studies, or concrete before/after outcomes to anchor the claims.

“ACV times number of deals equals revenue”

“you've got an answer in 30 seconds instead of the best case, 30 minutes”

The host brings a healthy sceptical framing and raises substantive structural challenges - lock-in, standardisation impossibility, keeping pace with change - but questions are extremely long-winded and grammatically difficult to follow, follow-ups rarely drill into David's specific claims, and no assertion goes meaningfully contested.

“I'm suspicious guy about the semantic layer and the reality of the semantic layer”

“I feel like the standardization is literally not possible. Like every vendor trying to push their one.”