Cellprose
An open plain-text format for spreadsheet workbooks. Every file is valid GitHub-Flavored Markdown, so workbooks render as tables and diff cleanly on GitHub, and a model can read them for fewer tokens than CSV. Formerly called Sheetmark.
Hi, I'm Utkarsh.
I build tools that help people and language models work with spreadsheets.
Products
Tablient
Coming soon.
karshlabs
The studio behind these products. Page coming later.
Cellprose in one look
A small model, stored as plain text. Formulas are written once per fill range, so a row of twenty formulas costs one line.
cellprose: 0.3
# Sheet: P&L
## A1:F4
#|A|B|C|D|E|F
-|-|-|-|-|-|-
1||Q1|Q2|Q3|Q4|Total
2|Units|1200|1344|1505.28|1685.91|5735.19
3|Price|45|45|45|45
4|Revenue|54000|60480|67737.6|75866.1|258084
```formulas
C2:E2 = B2*(1+Growth)
F2 = SUM(B2:E2)
B3:E3 = Assumptions!$B$3
B4:E4 = B2*B3
F4 = SUM(B4:E4)
```Research
How well do models answer questions about a workbook, and at what token cost, depending on how the workbook is written as text? Early results: Cellprose matched or beat CSV on accuracy for all seven models tested, cut output tokens by 16 to 31 percent, and cost about a quarter more input tokens.
About
Bio coming soon.