Memory by Contribution
How institutional memory gets built up by contribution rather than by ingestion — and why that distinction governs whether your AI compounds or commoditises.
Author Storefront · Ibex AI
Strategic writing on Context-IQ, governance, and sovereign intelligence.
Whitepapers and essays on how organisations turn context into a moat — and how to keep that moat theirs as frontier models converge. A running series, written from inside the work at Ibex.
The series
Featured whitepapers
Read the Context-IQ trilogy, or jump into the Scout/agent strategy paper.
I
Strategic Whitepaper · Foundation
The AI Moat for Organisations Is Context
Once frontier models converge, the moat is no longer model access — it's the institutional context an organisation puts around the model. Introduces Context-Based Knowledge Intelligence (Context-IQ) as a persistent, compounding intelligence layer.
Read paper
II
Strategic Whitepaper · Governance
Governing Enterprise Intelligence
Context-IQ creates intelligence. The governance layer makes that intelligence trustworthy enough to matter — covering trust, control, oversight, and responsible AI as architectural concerns rather than policy add-ons.
Read paper
III
Strategic Whitepaper · Sovereignty
Sovereign Intelligence
Using frontier models without surrendering institutional knowledge. Introduces a closed-loop architecture — CQ-MIE, the Semantic Airlock, and the mcp-midnight connector boundary — for organisations that need both frontier reasoning and full sovereignty over what leaves the building.
Read paper
IV
Strategic Whitepaper · Agentic AI
Beyond the Agent: Why Scouts Scale Agentic AI
Explains why one-off AI agents stall at scale, and why persistent Scouts with memory, shared learning, receipts, and governed autonomy become the real unit of compounding agentic AI.
Read paper
Coming next
In the pipeline
Drafts in motion — titles and framings may shift before publication.
No longer only drafts — the runtime is live. The substrate these papers describe now runs on the host: the governed-learning rails, the compounding-precedent layer (memory that earns permanence through reuse, not ingestion), and the settled-trust instrument are all deployed with their telemetry moving. The papers document a system that exists; the autonomy it governs is being earned tier by tier, shadow-first — with the value-settlement (x402) and media-authenticity model pieces still maturing behind their connector boundaries.
PACT — the Trust Plane
The trust contract between models, contexts, and the humans accountable for what they produce. How a Programmable Accountability Contract Trail makes governance enforceable at runtime.