Formas.ai Launches Cartesian, an AI 3D Modeling Tool from Mouse to City
On September 14, Formas.ai launched Cartesian, an AI 3D modeling tool billed as 'from a mouse to a city,' powered by its ASTRA engine. It is currently available as a free preview via a waitlist and targets architecture and product design workflows.
Confirmed
- Flexible input: models can be generated from natural-language conversation, photos, sketches, floor plans, or scans, with natural-language control over which parts to keep or modify.
- Editability: each object exists independently; editing one item (e.g., swapping a chair) does not affect the rest of the room.
- Precise geometry: objects carry real faces, edges, and solids that can be inspected, measured, and modified, outputting true solids with clean topology rather than polygon-mesh approximations, with a hierarchy of named elements.
- Users can 'walk into' generated scenes to inspect and manipulate them, and can cut and modify 3D objects.
- Free during the preview phase.
Why it matters
- Most AI 3D generation produces mesh approximations that are hard to measure or edit. Cartesian's emphasis on solid-level precise geometry and independently editable objects, if it holds up, would plug directly into professional architecture and product-design workflows rather than mere concept visualization.
- Unified cross-scale generation (from single objects to entire cities) could notably expand the boundaries of AI spatial intelligence. The product remains in preview/waitlist stage, so real-world quality and engineering usability await third-party validation.
2026-09-14 ~ 2026-09-15 · 5 related posts
Primary sources
- Formas Unveils CARTESIAN: Turn Photos and Sketches Into Editable, Precise 3D Models — carlosbannon · 2026-09-14
- [source] formas_ai launches CARTESIAN, a spatial AI that cuts and rebuilds anything from mouse to city — carlosbannon · 2026-09-14
3 near-duplicate retellings: carlosbannon · carlosbannon · carlosbannon