Why it exists
There is a specific, common, avoidable disaster: someone double-clicks a .tsv or a .csv, a spreadsheet opens it, and the file is quietly changed. Identifiers become dates. Leading zeros vanish. A column of long numbers turns into scientific notation. Nothing warns anybody, and the damage is only discovered downstream.
The alternatives were a plain text editor, where the separators are invisible and a hundred-column file is unreadable, or a paid localisation platform, which is a large answer to a small question.
So: a grid that understands what a table file is — whatever format it happens to be written in — in a browser tab, that gives the file back exactly as it found it.
The rules it is built on
Your file never leaves your device. Not as a promise about what happens to uploads — there is no upload at all. Reading, checking, repairing and saving all happen inside this tab. Disconnect from the internet once the page has loaded and everything keeps working, which is the simplest way to check.
The bytes you did not change do not change. Line endings, byte order mark, quoting style and trailing newline are detected on open and reproduced on export. Edit one cell in a file of 50,000 rows and exactly one line of that file changes.
Nothing is repaired without being shown first. Problems are reported with counts and locations, and the fixes are offered. Whitespace is drawn in the grid where it sits, because an invisible character you cannot see is the entire reason the file is broken.
Some things are deliberately not automated. Duplicate keys are found and pointed at, never resolved. Only you know which of the two is right, and a tool that guesses is worse than one that stops.
The editor stays out of the way. No animation, no reveals, no shimmer — data appears instantly or the tool feels broken. The personality lives on pages like this one.
How it holds up
Large files stay quick because the reading and checking happen out of the way of the screen. 50,000 rows open in well under a second, and the view never locks up while a file is being examined.
Only the rows you can actually see are drawn, so scrolling stays smooth however long the file is — and however wide. A hundred columns is not a different experience from three.
Every change is recorded in a form that can be played backwards, which is what makes undo unlimited rather than rationed. Repairing 52 cells at once is a single step in that history, and a single keystroke to take back.
Page views are counted, and that is the whole of it: which page, where the link came from, a country and a device type. It is done without cookies, and the identifier it groups a visit by is thrown away and regenerated every day, so there is nothing that can follow you from one day to the next or from here to anywhere else. What you open, what you type and what you export are not part of it and never leave your device.
What it will not become
Not a spreadsheet. There are no formulas, no charts and no pivot tables, and adding them would mean interpreting values — the exact behaviour this was built to avoid.
Accounts and syncing are not in it. If they are ever added, they will be optional and additive: the local, private, no-account path will keep working exactly as it does now.
Cost
It is free and stays free. The guides and conversion pages carry advertising, which covers the domain and hosting. The editor is kept clear of it: at most a slot in the margin on a wide screen, never over the grid and never next to a control.