Knowledge Lab
What I'm building and testing, in the open.
Research Lab studies patterns in what other people build. Knowledge Lab is where I build and test my own: independently, on public or synthetic data, with the status of each artifact labelled honestly rather than implied.
My day job at Kraken doesn't appear here as a case study. See Work for what I can say about it, and why.
4 ARTIFACTS · 2 IN BUILD · 1 SCOPED · 1 TRACKINGSTATUS DESCRIBES AVAILABILITY, NOT MATURITY: LIVE · IN BUILD · SCOPED · TRACKING
The grid queue isn't a capacity problem.
It's a knowledge problem.
Britain's grid connection queue has grown to 125GW, against peak demand of around 45GW. The usual explanation is infrastructure: not enough grid, not enough capacity, not enough investment. I wanted to test another hypothesis.
What if part of the bottleneck is semantic?
Across the organisations coordinating the same physical grid, connection capacity, settlement, flexibility and assets are represented through different data models, schemas and definitions. I built an interactive Knowledge Graph to make that fragmentation visible.
7 organisational nodes · relationship mapping · knowledge-debt model · live reconciliation-cost simulator
The 125GW/45GW figures and organisational fragmentation are grounded in public Ofgem/NESO material. The knowledge-debt scoring and cost model are an original illustrative analytical framework, not official industry figures.
KARI
Knowledge Architecture Readiness Index
AI-ready knowledge starts long before RAG.
KARI is a framework I've developed to assess whether an organisation's underlying knowledge architecture is ready to support humans, search, automated translation, RAG, MCP and AI agents. It looks beneath the interface at the things those systems depend upon.
The question isn't "Are you using AI?" It's:
Is the knowledge underneath it ready for AI?
Energy Flexibility Knowledge Graph
A knowledge graph modelling how distributed energy assets, flexibility services and APIs relate to one another, built entirely on public or synthetic data, not on anything from a current or former employer.
SCOPE: DER/VPP/flexibility-service/API relationships, public and synthetic data only, no employer material.
The point isn't to reproduce any specific product. It's to demonstrate knowledge-graph and ontology capability against a domain (energy flexibility) without exposing anyone's internal architecture.
Signal Stack Ontology
The evidence-classification model that underpins the Research Lab practice, expressed as an ontology: five signal classes and two relationships that keep observation and conclusion from collapsing into each other.
Used in every Research Lab edition to keep title signals, requirement signals and synthesis signals from being treated as the same strength of evidence.