Making deck.gl AI-Ready: What Seven Frontier Models Taught Us About Maps

At the Open Visualization Collaborator Summit at ETH Zürich, CARTO founder Javier de la Torre presented an evaluation of how seven advanced AI models recommend, generate, and interact with deck.gl maps, finding that while most can produce functional and often polished visualizations, their code commonly relies on outdated deck.gl versions, deprecated integration patterns, and stale official examples. The tests showed broad agreement that deck.gl is suited to large-scale data visualization, MapLibre to basemaps and React applications, and Leaflet to simple maps, but models frequently lacked awareness of deck.gl 9.4 and struggled with confusing MapLibre integration options, skewed-data styling, legends, and silent failures in declarative JSON specifications. CARTO proposed making a schema-backed @deck.gl/json v2 the primary agent interface, with validation, conversion reports, state read-back, patch editing, data-source support, and standardized color expressions, while also improving documentation through llms.txt files, Markdown pages, agent skills, and updated examples. The presentation additionally called for vis.gl to formalize AI-assisted contribution practices under the OpenJS policy and determine how automated development, review, verification, and security should operate as agents become both major users and contributors to deck.gl.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.