Every score Gridspace produces comes from an ontology — a structured theory of what makes a place work for storage. It is versioned, auditable, and open to argument. This is where we show it.
Browse the site types we screen for — what defines each one, why it works, what kills it, and how confident we are.
Explore the ontology →Every change to the scoring model, dated, explained, and attributed. What we changed, what evidence moved us, who validated it.
Read the log →The AI proposes site types and candidates. Humans confirm or reject. Both outcomes are published — including the misses.
See what we're testing →Disagree with our weights? Fork the ontology, change them, and run it against the real map. Publish your version.
Open the Lab →Why these weights? → The decision record that set them, including the recorded dissent.
Things we learned from running the ontology. Every finding has a number in it.
Six of nine failed on ownership fragmentation, not on grid or permitting.
Five categories, 23 weights, and the confidence rating that changes what a number means.
Adequate on geometry and access; fails on social license in every district tested.