Why we stopped building big spreadsheets
Sheets from us: 5 to 10 columns, 3 pages at most. Past 1,000 rows, connect by MCP.
The confession
Mark here. I never want to put data into a spreadsheet again.
I am happy to take your spreadsheets in. Building them is another matter. With data as complex as ours, it requires Maks, Mark or both crying themselves to sleep.
Neither Apps Script nor custom macros help. We have tried everything, from calling an LLM inside the sheet to running a database inside it. It is easier to build a password-protected custom app than to hand over a usable sheet.
But my sheets are usable
"Oh, but Mark, my sheets are usable," a bunch of people have said to me.
If your process is to shove the sheet into an LLM and get an output, or 100 drafted emails, you do not have usability. You have token increases.
We do our best never to give a person something ugly or unusable. A sheet we deliver holds interactions we code by hand so a human can use it. It is not a dump of data for an LLM. It has formulas, tables, API calls and more.
What we noticed this week
This week we thought it through. We had been building an app in Google Sheets for every customer: a database pushed into a sheet, then API and MCP connectors, then custom interface pieces, every time.
That is a lot of engineering to end up with a spreadsheet.
What we build instead
So these days we turn the spreadsheet your team lives in into a small site, made for how you work: an appeal tracker, a volunteer roster, a sponsor board, a list of parishes.
Same data. Fewer tabs. Nobody scrolling to column AF.
The new rule
5 to 10 columns, 3 pages.
At most, from now on. If you want a sheet from us, we can give it to you or help you export it.
Over 1,000 rows, I will ask why.
I would question where the export is going and what it is for.
Connect instead.
Connect by MCP to our whole dataset. It is more accurate, up to date and reusable. A snapshot is static, and our database grows daily.

