From the Credible team
Practical thinking for teams building AI, analytics, and data products that people can trust.
Engineering
17 MIN READ
Your business logic lives in LookML, and only Looker can read it. Here's how to move it into a governed model you own - and build something better than the dashboards you're leaving behind.
Oliver Larsson
Solutions Engineer @ Credible
Engineering
10 MIN READThe modern data stack set the order we model in a decade ago, for consumers who no longer dominate: pipeline first, meaning last. Flip it. Write down what your data means first, in one language -- and let the engine build the transformations, tables, governance, and context underneath it.
Oliver Larsson
Solutions Engineer @ Credible
Engineering
22 MIN READThe AI Analytics Engine is not another data platform. It is a paradigm shift: software moved from hand-written assembly to compilers to managed runtimes -- and that same move is now happening in data. You write down what your data means as a data model, in one language. The engine generates the rest: dashboards and data apps, the semantic layer that serves them, the pipelines and optimized storage underneath, and governance on every query.
Kyle Nesbit
CEO & Founder @ Credible
AI & ML
13 MIN READA detailed comparison between the Credible and dbt semantic layers, drawn from two production agentic-analytics deployments and organized around four questions to ask of any solution: what can an agent ask, what can it know, what can't it do, and how does the platform run.
Adam Ribaudo
Founder, Noise to Signal
Open Source
14 MIN READThe whole series, run end to end on one messy, real dataset: 320,000 public security vulnerabilities across six overlapping feeds and severity scales that disagree. An agent builds the model, the definitions get locked, a loaded question gets a defensible answer, and a data app ships it.
Ofer Mendelevitch
DevRel @ Credible
Engineering
15 MIN READThe dashboard was never the problem -- the canvas was. When an agent with the right skills hand-authors the HTML against a governed model, a dashboard stops being a config blob and becomes source code: reviewable, versioned, testable, and shipped like the rest of your software.
Nathan Huff
Head of AI & Application Development @ Credible
Open Source
10 MIN READThe open-source analysis skills encode the discipline that separates an analyst from a confident guesser: resolve words into definitions, ground the scope, verify before presenting. We walk one real question through it, checks and all.
Oliver Larsson
Solutions Engineer @ Credible
Open Source
13 MIN READWe open-sourced the Malloy modeling skills that teach an agent how to investigate data before it models it: prove grain and joins with queries, flag the business decisions the data can't answer, document the result, and make its assumptions visible.
James Swirhun
Head of Product @ Credible
Open Source
7 MIN READMalloy Publisher now ships five MCP tools, thirty agent skills, and the design principles that decide what belongs in each.
Monty Lennie
Software Engineer @ Credible
Open Source
5 MIN READA few-minute setup: point your agent at Malloy Publisher and start asking real questions of your own data.
Monty Lennie
Software Engineer @ Credible
Open Source
9 MIN READThe agent skills and MCP tools behind Credible are now open source, in Malloy Publisher. Excellence is no longer a moat — so we gave the layer away and bet on trust, durability, and distribution instead.
Kyle Nesbit
CEO & Founder @ Credible
AI & ML
15 MIN READAtlas lets anyone explore a catalog of public datasets by asking questions in plain English. Built on Credible and Malloy, it turns natural-language questions into governed, interactive charts — and full data stories that you can publish, all in one place.
Girish Jeswani
Software Engineer @ Credible
AI & ML
5 MIN READCharting is commoditized. Meaning is the product — why the data model is what analysts, engineers, and AI agents all build on.
Kyle Nesbit
CEO & Founder @ Credible
AI & ML
13 MIN READAI agents are learning to read documents; the harder problem is using the structured data that runs the business. Here's how Malloy turns a complex healthcare schema (OMOP) into a governed model agents can query reliably — and how Credible serves it to production agents over MCP.
Ofer Mendelevitch
DevRel @ Credible
AI & ML
7 MIN READData warehouses can now run embeddings, classification, and LLM inference natively. Anyone can build an ML pipeline — but without evaluation, governed logic, and version control, you don't know what you're getting. Malloy brings structure to warehouse-native ML.
James Swirhun
Head of Product @ Credible
AI & ML
7 MIN READEntity matching pipelines built on string-matching heuristics are brittle, expensive to maintain, and impossible to scale. Embedding models dramatically outperform them — but only if you can evaluate and iterate on results. Here's how to build production-grade entity matching you can actually trust.
James Swirhun
Head of Product @ Credible
Engineering
6 MIN READdbt brought software engineering to SQL, but SQL + Jinja + YAML complexity compounds at scale. Malloy unifies transformation, modeling, and materialization in one declarative language — type-safe, composable, and AI-ready.
James Swirhun
Head of Product @ Credible