Math, at scale.
Grids is a cloud calculation platform. Data keeps the familiar spreadsheet workspace. Grid 0.66.8 adds Code for source authoring, App for focused model views, and Chat for conversation. All four layers use the same model and calculations; availability follows your workspace runtime version.
- 01
Formulas are functions
Cells are variables. Formulas are functions. A workbook is a program — you have been writing them since seventh grade, you just never had a runtime that took them seriously. Grids is the runtime.
- 02
The grid is the engine
Behind every cell is a calculation pool. Formulas fan out across machines, run in parallel, and land back in your sheet as if nothing happened. A million-scenario Monte Carlo, a warehouse-scale join, an exact symbolic solve — the sheet doesn’t know it’s a supercomputer.
- 03
The math doesn’t stop
A hundred-thousand rows. A hundred-million scenarios. The math grows; the grid grows with it. There is no “too big” — only longer or shorter waits, and rarely longer.
You write the math. The grid runs it. Type a formula, press enter, the cloud answers.
Four ways to go deeper.
Each page is one slice of how the conceit becomes real — a different scenario, a different surface of the same product.
A spreadsheet that knows what it's saying.
Type tags, spilling arrays, pipelines, and 2,903 built-in functions — the surface area, with a hiring rubric to ground it.
Source on the left. A running model on the right.
Grid source → Rust compiler → LIR runtime. One calculation path serves the model across its different views.
A spreadsheet that fetches the world.
Async built-ins, imported APIs, native SQL with warehouse pushdown, the dirty → ready lifecycle, and one operator that decides whether work is eager or lazy.
Describe the problem. Inspect the proposed model.
Chat for working with a model, HTTP endpoints for generation and explanation, and MCP tools for external agents. Review the diagnostics and proposed changes before applying them.