Context and Memory Infrastructure
Research question
What should a system remember, and what should it be able to prove about what it remembered?
Working thesis
Context is infrastructure, not prompt stuffing. Retrieval conditioned on the task, explicit budgets, pruning, provenance, and access boundaries are engineering concerns with their own evaluation criteria — and institutional memory is a storage substrate problem before it is a model problem.
A thesis is the lab’s current position, not a result. What supports it is listed below; what remains unresolved is stated at the end of this page.
Publications
3 attributedFrom Context Engineering to Hierarchical Engineering
Toward Layered Cognitive Infrastructure for Agentic Systems
Research ReleasePublishedApril 3, 2026Context Is Infrastructure, Not Prompt Stuffing
Why retrieval engines must evolve to be task-aware.
Architecture NotePublishedNovember 17, 2025StrongHold and Governed Data Archives
Why research data needs an archive layer, not a notebook.
Technical NotePublishedAugust 17, 2025
Apparatus
3 systemsSystems attached to this program are what the lab builds to investigate the question. Maturity, evidence level, and current activity are recorded separately on each system — a system attached here is not thereby claimed to be production-ready.
StrongHold
Adaptive Storage & Archive Intelligence
A smart storage system for adaptive compression, archival intelligence, retrieval, and restoration — storage that understands artifact structure and can explain its own strategy decisions.
Artchiv
Creative Media Archive & Provenance Platform
The source of truth for creative media — a governed archive for versioned assets, provenance records, retrieval, and inspectable media lineage across studios, agencies, and production workflows.
ACE
Context Engine
A context engine and retrieval infrastructure for task-conditioned retrieval, source ranking and pruning, context budgets, and citation-preserving assembly across AI workspaces and agent workflows.
Open questions
Published as open. These are the parts of the program the lab does not consider settled.
- 01
How is context quality evaluated independently of the model consuming it?
- 02
What is the right unit of institutional memory — the document, the trace, or the decision?
- 03
How do access boundaries survive retrieval across a whole organization's memory?