Lab
Experiments
& tools.
Things that don't belong in /projects — explorations, patterns, work-in-progress.
- ongoing
Agent orchestration patterns
AI infrastructure
Nodus internals
Patterns for reliable multi-agent coordination: checkout locks, run provenance, incremental context. The operational primitives that make 17-agent systems trustworthy at daily-driver scale.
- drafting
Design token ontology
Design formalization
Structured Thinking
A formal vocabulary for design tokens that goes above W3C DTCG syntax — semantic aliasing, inheritance rules, and cross-brand resolution without losing the source of truth.
- active
Finance UX density patterns
Finance domain
Design systems
The layout conventions for complex financial interfaces: 32px compact rows, tabular numerics, status indicators that scan at a glance. Things that don't exist in standard design systems but are required in treasury and capital markets.
- shipped
Vector knowledge at personal scale
AI infrastructure
Knowledge systems
630K+ chunks vectorized for 5ms search. What actually works for personal knowledge management at practitioner scale — chunking strategies, embedding models, retrieval tuning. Nodus internals.
- drafting
Agentic UX trust patterns
AI-native design
UX patterns
How do you design interfaces for AI systems that take real actions? Provenance display, confidence signals, reversibility affordances, handoff UX. The patterns that make agentic systems usable by non-engineers.
These are working notes, not finished products. Some become Structured Thinking articles. Some become Nodus features. Some stay here.