Equally Powerful in the Semantic Layer
Summary: Paul’s observation, made in a conversation about this wiki’s own creation, that generative AI has flattened the syntactic barrier between semantic experts (who understand the problem) and syntactic experts (who translate it into working systems), so semantic experts can now achieve syntactic success directly.
Sources: conversation with Paul (2026-07-08)
Last updated: 2026-07-08
Paul’s own framing: “We are all equally powerful in the Semantic Layer.” (source: conversation with Paul, 2026-07-08)
The old divide
Historically, the role of semantic expert, the person who understands the business problem, the domain, what actually needs to happen, was distinct from the role of syntactic expert, the person who could translate that understanding into working code or systems. A large amount of effort went into translating across that divide, and the translation was hard precisely because semantic and syntactic experts often lacked a strong shared vocabulary and world model (source: conversation with Paul, 2026-07-08).
What generative AI changes
With easy access to foundation models, semantic experts can now express their needs directly through AI interaction and achieve syntactic success themselves, with much less overt dependence on a specialized syntactic partner who would otherwise need extensive grounding to be useful. The AI absorbs a large share of the translation work that used to require a dedicated human intermediary (source: conversation with Paul, 2026-07-08).
The “equally powerful” claim is specifically about the semantic layer: everyone, regardless of coding background, now has meaningfully similar access to syntactic execution through AI. What still varies, and what now matters more rather than less, is the quality and depth of the semantic understanding someone brings to the interaction.
Where this connects in the wiki
This sharpens an observation already in context engineering: “the barrier to creating an AI-based solution has come down… many people, not just a small specialized group of data scientists, now have the tools to leverage pre-trained foundation models.” The Semantic Layer framing names precisely what didn’t get easier even as that barrier fell: semantic depth, not syntactic skill, is now the scarce resource.
It also reframes the asset-ownership role. Asset Owners were described as “translators at the boundary between practitioners and users.” If AI absorbs much of that syntactic translation work, the Asset Owner’s value shifts even further toward the semantic side: understanding what users actually need, judging whether an AI’s syntactic output actually satisfies that need, and knowing when to distrust a fluent-sounding answer.
This is also the flip side of foundation-of-trust’s claim that trust comes from judgment, not skill. If syntactic skill is being equalized by AI, judgment, knowing what’s actually needed, recognizing when an AI’s output is wrong despite sounding right, and knowing when to ask for help, becomes the harder-to-automate differentiator.
This same line has a first documented origin outside of conversation: information advantage, Paul’s own framework for insurance-industry competitive advantage, lands on identical language when describing what happens to the “Knowing More” advantage under generative and agentic AI (source: INFORMATION ADVANTAGE.md).