Open-source computational tools built alongside the artwork practice. Each tool grew from a concrete research question — and each generates data that feeds back into the work. All tools run locally, require no cloud services, and are free to use.
Mapping the unnamed gaps in language
A navigable terrain built from 36,125 concepts in 384-dimensional embedding space. Finds and measures the “deserts” — regions of meaning with no syntactic representation in natural language. Companion to the oil painting Are there deserts in vector space?
View detailsStandardised quality reports for dimensionality reduction
Takes high-dimensional data and its low-dimensional projection and produces a one-page quality report: trustworthiness, continuity, neighbourhood preservation, Shepard goodness, normalised stress, and silhouette score. Answers the question: how much should you trust your map?
View detailsFrom dataset to physically buildable sculpture design
Reads a CSV dataset, extracts its relational structure using pairwise similarity or explicit edge lists, and produces a PDF fabrication document: node coordinates, a rod cut list, materials summary, and build notes. The sculpture IS the data — no manual position overrides permitted.
View detailsFind the deserts in any conceptual space
Feed it a corpus of descriptions from any categorisable domain and it maps the semantic space those concepts occupy, identifies the low-density regions — the deserts — and generates ranked candidate descriptions for what could inhabit those gaps. Not a search engine. Not a recommendation system. A strategic positioning instrument.
View detailsScore every trust-boundary crossing in your infrastructure
Reads an architecture description — YAML, Terraform state, Docker Compose, or Kubernetes manifests — and scores every communication edge using a trust-boundary thinness formula. Produces a self-contained HTML report with an interactive graph, a one-page client deliverable, and a JSON export. Runs entirely offline in under a minute.
View detailsSecurity and conformance testing for MCP servers
Automated conformance and security testing against the Model Context Protocol (MCP) specification 2025-11-25. Tests initialization handshake, capability declarations, tool invocation, JSON-RPC error handling, shell injection, input validation bypass, and information disclosure. Produces text or JSON reports.
View details