Magic vs Science & Introduction to OpenMOCA

Room: Yard Bar (Ground Floor next to the restaurant)

1830 - Arrive
1850 - Intro
1900 - Introduction to OpenMOCA: Portable grounding for AI
1945 - Break
2000 - Magic vs Science: Making AI Coding Agents Prove Their Work
2045 - Finish

Magic vs Science: Making AI Coding Agents Prove Their Work
This talk moves beyond prompt tricks, skills and the latest witchcraft to show how engineering practices can make generated code prove itself. We'll cover and demo the steps you can take to ensure you build quality, maintainable, and observable solutions when coding with AI.

Jag Reehal
https://www.linkedin.com/in/jag-reehal/

Highly recommended Principal AI Engineer based in Cambridge, with certifications in AI, & AWS with over 20 years of experience architecting scalable front-end and back-end systems for both startups and large enterprises.

Introduction to OpenMOCA: Portable grounding for AI
Most AI systems answer from knowledge locked inside one product: a pipeline, a vector store, an embedding model. You can't easily move that knowledge, check an answer against its source, or prove the content hasn't changed.

OpenMOCA is an open standard for packaging that knowledge. A MOCA package holds the content an AI answers from, how it's organised, and the original sources. A single digest seals all three, and the package can be signed. Any conforming Reader can open it, verify it and search it the same way, on a server or on a laptop with no network, whatever model or framework you use.

In this talk we'll look at:
- Why MOCA: grounding you can move between systems\, cite and check\, instead of rebuilding a RAG pipeline for every product.
- What's in a package: Markdown content\, sources and evidence\, an optional ontology\, and sidecar search indexes.
- The Readers: TypeScript\, .NET and Python. Each one does keyword\, semantic and hybrid search through one call\, and every result cites the exact passage it came from.
- The tools: converters from folders\, PDFs and YouTube; lint\, sign and index tools; and an MCP server for agents.
- Use cases: manuals and procedures\, regulation\, vendor docs\, support knowledge and recorded talks.
- A live demo: "Try OpenMOCA". We'll search the Kubernetes docs and the UK Data Protection Act with no AI model at all\, compare keyword\, semantic and hybrid results side by side\, follow an ontology from a topic to what it builds on\, then chat with a package where every answer is cited and opens at its source.

You'll leave knowing how to package your own content and how to put a Reader into your app.

Pete Cleary
https://www.linkedin.com/in/pete-cleary-8628084
I’m a software architect and engineer with over 25 years’ experience building digital products. I’ve founded and run two successful technology companies and have also worked for Visa Inc. These days, I’m focused on the opportunities and challenges that AI brings, particularly around building useful, practical systems and understanding how we can apply the technology responsibly. I’m particularly interested in how we can use AI for the greater good in society.