The Semantic Intelligence Platform Pronounced /koʊˈdʒɛnɪti/ or in simpler terms: koh-JEN-uh-teehttps://bsky.app/profile/coginiti.cohttps://bsky.app/profile/coginiti.co/post/3mwbjo2wdmw2lLast few hours at Big Data LDN. Stand E34 until close. If you caught either talk and want to see the ideas running rather than described, there's a live environment on the stand and no queue at this hour. #BigDataLDN24 Sep 2026 15:03 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mwbjo2wdmw2lhttps://bsky.app/profile/coginiti.co/post/3mwbaqgsp6y2d"Semantic layers are too rigid for AI" The opposite. The more probabilistic the interface, the more it needs a deterministic contract underneath. Without one, the same question gives different answers by prompt wording or model version. https://coginiti.co/blog/semantic-layers-agents-fud/24 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mwbaqgsp6y2dhttps://bsky.app/profile/coginiti.co/post/3mwaw6pbh7t2qDay two, 25 minutes out. AI Context Theatre, 10:40. Richard Hall on what semantic models, ontologies and context engineering each actually do, and when to reach for which. Stand E34 after. #BigDataLDN24 Sep 2026 09:15 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mwaw6pbh7t2qhttps://bsky.app/profile/coginiti.co/post/3mw6fpnbijb2zStarting in 25 minutes. Lightning Talks Theatre, 10:40. Matthew Mullins on context, governance and cost control for AI agents. Stand E34 after. #BigDataLDN23 Sep 2026 09:15 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mw6fpnbijb2zhttps://bsky.app/profile/coginiti.co/post/3mw47slny762s"Nobody agrees on the definitions anyway." That's the point. Disagreement is the reason semantic layers exist. Without one, four teams' definitions of "customer" get encoded silently into dashboards and queries. With one, disagreement becomes a governed decision. https://coginiti.co/blog/semantic-layers-agents-fud/22 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mw47slny762shttps://bsky.app/profile/coginiti.co/post/3mvzuvolxay25Big Data LDN this week. Stand E34, both days. Wed 23rd, 10:40, Lightning Talks Theatre: Matthew Mullins, Not Your Mum's Semantic Layer. Thu 24th, 10:40, AI Context Theatre: Richard Hall, The Context Stack. Still time to register free. #BigDataLDN https://www.bigdataldn.com/en-gb/register.html?code=1718125174025894-MKP21 Sep 2026 14:03 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mvzuvolxay25https://bsky.app/profile/coginiti.co/post/3mvpwdva6al2zNobody budgets for the tokens an agent spends finding its way around. Lists tables. Samples rows. Tries a join, fails, tries another. Then rediscovers the whole schema on the next question. Token spend is an architecture problem. #BigDataLDN #FinOps17 Sep 2026 15:02 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mvpwdva6al2zhttps://bsky.app/profile/coginiti.co/post/3mvpnhzth6n2b"Semantic layers are only useful after the data is perfect." The perfection trap. Nobody has perfect data, and waiting for it means waiting forever. The semantic layer is part of how you make data usable. Start with the metrics that matter most, not day-one everything. https://coginiti.co/blog/semantic-layers-agents-fud/17 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mvpnhzth6n2bhttps://bsky.app/profile/coginiti.co/post/3mvkvhugqxj2sThe failure mode of an agent on raw schemas isn't that it breaks. It's that it succeeds. Counterintuitive right? The agent returns some number. Correctly formatted, plausible, wrong, because it picked orders.amount and you define revenue net of returns. Nobody catches it. #BigDataLDN15 Sep 2026 15:03 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mvkvhugqxj2shttps://bsky.app/profile/coginiti.co/post/3mvkmk7xr3y2s"We can solve this with prompt engineering." A prompt that says "use approved revenue" is not a governed revenue measure enforced in the execution path. Prompts are hard to audit, version, and test. Prompts guide behavior. A semantic layer governs meaning. https://coginiti.co/blog/semantic-layers-agents-fud/15 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mvkmk7xr3y2shttps://bsky.app/profile/coginiti.co/post/3mv6d6a37cs26An analyst joining your company gets onboarding and a colleague to ask. An agent gets a connection string. No institutional memory, nobody to ask, machine speed. When it guesses wrong about revenue it does so confidently and at scale. #BigDataLDN #AIAgents10 Sep 2026 15:04 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mv6d6a37cs26https://bsky.app/profile/coginiti.co/post/3mv627lbruy2b"Governance kills agent autonomy." A self-driving car still needs traffic laws and lane markings. Let agents explore freely. But when one answers an enterprise metric question, it shouldn't invent the definition. Autonomy to explore, governed semantics to answer. https://coginiti.co/blog/semantic-layers-agents-fud/10 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mv627lbruy2bhttps://bsky.app/profile/coginiti.co/post/3mv3n63gars2uAgents stopped failing at SQL. For the strongest model tested, 60%+ of remaining failures were semantic: wrong metric, wrong source, wrong join key. New research on what survives as models absorb the stack. https://coginiti.co/blog/model-eats-the-stack-semantic-context/09 Sep 2026 13:25 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mv3n63gars2uhttps://bsky.app/profile/coginiti.co/post/3muzc5oojyg2nSemantic model. Ontology. Context engineering. Three terms used interchangeably by people selling three different things. Richard Hall sorts them out at Big Data LDN. AI Context Theatre, Thu 24 Sept, 10:40. What each one does, and when it doesn't. #BigDataLDN08 Sep 2026 15:02 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3muzc5oojyg2nhttps://bsky.app/profile/coginiti.co/post/3muyzbqu3bb2d"It's just an old BI idea with new marketing." The idea is decades old. That's a strength. What changed is the number of consumers: agents, notebooks, workflows, embedded apps, decision systems. AI multiplied the systems that need governed access to meaning. https://coginiti.co/blog/semantic-layers-agents-fud/08 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3muyzbqu3bb2dhttps://bsky.app/profile/coginiti.co/post/3mumpt6lf4t2mFor 20 years the semantic layer was a BI convenience. Friendly names so nobody had to remember which table held revenue. That was built for humans clicking dashboards. The user has changed. Matthew Mullins, Big Data LDN, Wed 23 Sept 10:40, Lightning Talks Theatre. #BigDataLDN03 Sep 2026 15:02 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mumpt6lf4t2mhttps://bsky.app/profile/coginiti.co/post/3mumgx5j75g2r"Semantic layers hide the underlying data." A good one does the opposite: exposes definitions, shows the generated SQL, provides lineage, keeps version history When an agent answers, you should see the metric, the filters, and the query it ran. Explicit and inspectable https://coginiti.co/blog/semantic-layers-agents-fud/03 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mumgx5j75g2rhttps://bsky.app/profile/coginiti.co/post/3muhovt6cpd26We're exhibiting at Big Data LDN 2026. 23-24 Sept, Olympia London, Stand E34. Two talks: Matthew Mullins on the semantic layer as the control point for AI agents (Wed), Richard Hall on the context stack (Thu). Free tickets below. #BigDataLDN01 Sep 2026 15:03 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3muhovt6cpd26https://bsky.app/profile/coginiti.co/post/3muhfzckvjn2x"Semantic layers are just another abstraction." Everything in analytics is an abstraction. Tables abstract files. Views abstract queries. SQL abstracts execution. The question is whether it captures the right contract. A semantic layer abstracts business meaning. https://coginiti.co/blog/semantic-layers-agents-fud/01 Sep 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3muhfzckvjn2xhttps://bsky.app/profile/coginiti.co/post/3mtvsqsr7qt27"The LLM can just retrieve the documentation." Docs explains revenue in prose. The model still has to translate prose into query logic, and that translation is where drift enters: right definition, wrong table, wrong grain. https://coginiti.co/blog/semantic-layers-agents-fud/25 Aug 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mtvsqsr7qt27https://bsky.app/profile/coginiti.co/post/3mtjag7h2d62k"Ontologies and knowledge graphs will replace semantic layers." They describe how the business thinks about itself. Powerful, and complementary. But an ontology describes the business. A semantic layer has to answer questions against the data. https://coginiti.co/blog/semantic-layers-agents-fud/20 Aug 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mtjag7h2d62khttps://bsky.app/profile/coginiti.co/post/3mte7iezw262v"Semantic layers are only for business users." A leftover from the self-service BI era. Analysts, data engineers, analytics engineers, data scientists, applications, and agents all consume analytical meaning. Semantic layers are for every consumer, including software. https://coginiti.co/blog/semantic-layers-agents-fud/18 Aug 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mte7iezw262vhttps://bsky.app/profile/coginiti.co/post/3msxn5s5qns2j"Semantic layers are too hard to maintain." You're already maintaining semantic logic, scattered across snippets, stored procs, notebooks, and Slack. The layer doesn't add the burden. It makes it visible. Maintain it explicitly, or let it sprawl. https://coginiti.co/blog/semantic-layers-agents-fud/13 Aug 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3msxn5s5qns2jhttps://bsky.app/profile/coginiti.co/post/3mssmj64kha23"Semantic layers are brittle because the business changes." The business changing is the reason to govern definitions, not skip them. The alternative to governed change is silent drift across queries, dashboards, notebooks, and prompts. Pick which you'd rather have. https://coginiti.co/blog/semantic-layers-agents-fud/11 Aug 2026 12:29 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mssmj64kha23https://bsky.app/profile/coginiti.co/post/3msayxi5taa2o"Semantic layers can't handle complex analytics." They aren't meant to replace SQL, modeling, or exploration. But complex work still needs stable definitions: which customers are active, what counts as revenue. Not a cage. The governed foundation you build on. https://coginiti.co/blog/semantic-layers-agents-fud/04 Aug 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3msayxi5taa2ohttps://bsky.app/profile/coginiti.co/post/3mrugmvmdvm2q"Semantic layers are just metric stores." A metric without its business structure is incomplete. "Revenue" by what? Customer? Region? Recognized date? At what grain? Through which join path? Semantic layers model the structure, not just the formula. https://coginiti.co/blog/semantic-layers-agents-fud/30 Jul 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mrugmvmdvm2qhttps://bsky.app/profile/coginiti.co/post/3mrpfp3gkab2i"Semantic layers slow analysts down." Bad governance does. A good semantic layer speeds people up: no rediscovering the approved revenue calc, no reverse engineering joins, no asking five people which churn is correct. Restriction slows. Reuse accelerates. https://coginiti.co/blog/semantic-layers-agents-fud/28 Jul 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mrpfp3gkab2ihttps://bsky.app/profile/coginiti.co/post/3mrctegjuj32v"BI tools already have semantic layers." They do, and they were built to govern dashboards inside one tool. Governed meaning now has to serve SQL analysts, data products, APIs, and agents. BI semantics can't reach all of that. Enterprise semantics have to. https://coginiti.co/blog/semantic-layers-agents-fud/23 Jul 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mrctegjuj32vhttps://bsky.app/profile/coginiti.co/post/3mr5sgma7nf2z"The warehouse should own the semantic layer." Fine if you live inside one platform. Most enterprises run many, plus BI tools, notebooks, APIs, and agents. Trap semantics in one warehouse and consistency becomes lock-in. Enterprise meaning has to be portable. https://coginiti.co/blog/semantic-layers-agents-fud/21 Jul 2026 12:24 +0000at://did:plc:giu5o7uzub4qsjdxjipiizem/app.bsky.feed.post/3mr5sgma7nf2z