Topic hub
Epistemic Engineering
How do systems turn evidence into decisions that deserve to be trusted?
A working theory of knowledge infrastructure: representation, retrieval, synthesis, evaluation, assurance, and the boundaries where uncertain work becomes an accepted result.
Research threads
03What can deterministic machinery genuinely establish, and which uncertainties remain outside its tests, specifications, and repeatable outputs?
openHow do checks and balances make human and AI decisions trustworthy?What institutional machinery lets an organization delegate reasoning while preserving evidence, authority, accountability, and the ability to intervene?
openHow can structured disagreement lead to better decisions?How should systems preserve competing interpretations long enough to test them, without turning every decision into an endless debate?
Writing
04Working note
A Unified Theory of Epistemic Engineering
A four-dimensional theory of AI system difficulty, with practical machinery for designing and diagnosing production systems
Draft
Deterministically Wrong
What deterministic software can prove, what tests cannot see, and why assurance is ultimately a domain decision
Published essay
The Shadow of Compression
On symbolic and statistical learning as two directions toward the same fixed point
Skeleton
Some Systems Are Slow Because Trust Is Slow
On why accelerating knowledge work does not remove the institutional work required to make decisions trustworthy
Code and models
03Cataloguing
Practical AI Catalogue
Mapping deployable multimodal AI capabilities to real business functions, constraints, and evaluation methods.
Research notes
Information Retrieval as Epistemic Architecture
A research notebook about representation, bounded attention, and the systems that make knowledge retrievable.
Research · Open source
Latent Topologies
Research into the geometry of language-model representations using persistent homology and Hodge decomposition.
Still unresolved
Open questions
- Which parts of knowledge work should be deterministic, and which must remain exploratory?
- How can an AI system preserve the evidence and assumptions behind a conclusion?
- When is aggregate performance insufficient and individual decisions must be defensible?