Diagnostic Engine: Registry-Driven Knee Symptom Reasoning
A technical note on a knee-only reasoning engine with registry-defined evidence, safety escalation outside ranking, and an inspectable session ledger.
Writing
Technical writeups and notes.
Technical notes, architecture writeups, and blog posts.
Latest
Deterministic Compilation for Structured LLM Workflows
A long-form breakdown of typed node registries, static validation, and why planning should be constrained before codegen begins.
Governed Memory Layer for Multi-Agent Shared State
A systems-first writeup on governed memory promotion, append-only events, canonical projections, pending ambiguity, and why shared state should be earned rather than appended.
Open writeup →Blogs
The Basilisk Is the Race, Not the Model.
On AI as an incentive race, not a single monster, and why governed acceleration is the real architecture problem.
Read blog →Writing updates
Follow the blog.
Get new architecture notes on governed AI systems, memory, reliability, and deterministic execution.
Technical Writeups
Technical writeups.
Longer system notes and architecture breakdowns.
Diagnostic Engine: Registry-Driven Knee Symptom Reasoning
How a bounded knee symptom engine uses registries, safety-first escalation, deterministic scoring, and persistent session state to stay narrow and inspectable.
Deterministic Compilation for Structured LLM Workflows
A writeup on moving from free-form workflow generation toward typed planning, compile-time checks, and deterministic code output.
Governed Memory Layer for Multi-Agent Shared State
How private notes become governed canonical memory through interpretation, resolution, validation, deterministic projection, and pending-state deferral.
Read now →