What Olbo does not do

It can miss things

The detectors look for shapes, and for the words around them. A card number has to pass the Luhn checksum, an IBAN has to satisfy mod 97, a US routing number has to satisfy the ABA weighting, and a bank account number is only taken when something beside it says what it is. That is deliberate. A detector that guesses covers the quantity column in your spreadsheet and calls it a phone number.

The cost is the other direction. Something written in a way the detectors do not recognise goes through untouched. Read what comes out before you send it on. Every match is shown where it sits so that you can, and anything missed can be highlighted and covered by hand.

Names are off when you arrive

A name has no shape to check. There is no arithmetic that says whether a word is a person, the way there is for a card number, so the detector gets it wrong in both directions: it covers a product name you needed, or it misses a name and you send the file believing it was caught.

Tick Names if there are people in the file and you want it to try anyway, and check that result more carefully than the rest.

A scan has nothing to read, and there is no OCR

A scanned document is a picture of a page. There is no text in it to search, so no detector can find anything, and Olbo says so rather than reporting zero matches and letting you believe the page came back clean.

It will not read the words out of the picture for you. Every OCR engine that runs in a browser is WebAssembly, and Olbo ships no WebAssembly and leaves it out of its content security policy. That is a trade rather than an oversight: the policy is part of what makes the promise about your files something you can check rather than something you have to believe. On a scan you draw the boxes yourself, and what you draw is burnt into the picture when you save.

Saving a Word file writes a new one

A .docx is written fresh from what is on the page rather than edited inside the file you opened. Headings, bold and italic, lists, tables, links and pictures come through both ways.

Headers, footers, footnotes, comments, tracked changes, page setup and text boxes do not. If the document you opened had any of those in it, keep the original.

Sheets is not Excel

Fifty-two functions, which covers most of what people actually type: SUM, AVERAGE, COUNT, COUNTIF, SUMIF, VLOOKUP, IF, IFERROR, ROUND, STDEV.S and VAR.S, and the text ones like TRIM, LEFT and CONCAT.

Charts are drawn from what you select and saved as a PNG or SVG picture: they are not written into the workbook or the CSV. Pivot tables, macros and conditional formatting are not there. A workbook opens with its own formulas intact, and the parts Olbo does not touch, its own charts included, are written back as they were, so a file that uses them is not damaged by being opened here.

Where a file can be saved depends on the browser

Saving straight back into the folder you opened from needs the File System Access API. Chrome and Edge have it. Safari and Firefox do not, and in those a save is an ordinary download to wherever your browser puts downloads.

That works, and it is not the same thing. If you are editing in place from a folder, use a browser that supports it.

In Chrome and Edge on a computer, a file over about 400 MB is written straight into a place you choose instead of going to your downloads, so the browser asks where it should go first.

A licence is a browser, not an account

There are no accounts, so your key lives in the browser you pasted it into. It does not follow you to another machine, another browser, or a private window. Working on two computers means pasting the same key twice, which is expected and fine.

Clearing your site data clears the key along with everything else. Keep the email it came in.

A photo can be too big in pixels, not only in megabytes

What a picture costs to open is its pixels, decoded, rather than the file on disk, so pixels have a limit of their own. On a phone or tablet, Pictures stop at about 16.8 million pixels, which is 4096 by 4096, so a 48 megapixel phone photo of a few megabytes is refused there. On a computer they go up to about 268 million, 16384 by 16384, the most a canvas holds in Chrome and Safari. Docs, and the PDF tool when a picture becomes a page, stop at 16.8 million everywhere.

Pro does not raise either one. On a phone the limit is the memory a tab gets before the system closes it, on a computer it is the canvas, and a licence changes neither. Scale the photo down in any viewer and open it again.

How big a file can be

Free opens files up to 2 MB, everywhere. Pro has no size limit on a computer, in the desktop app or a browser, and opens files up to 50 MB on a phone or tablet. A phone is what breaks first: past that size its browser stops being slow and starts failing outright, with nothing to catch and nothing to report. In a browser, Olbo counts the device as a phone or tablet whenever anything about it says so, including a touch screen on a Linux laptop and a reported 4 GB of memory or less, and whenever it cannot tell. The desktop app counts as a computer.

Everything runs on your own processor, so how quickly a large file opens is a question about your machine. There is no queue behind somebody else's upload, and no server to make it faster either.

On a computer, the machine sets the limit

Pro has no size limit on a computer, so what stops a big file there is its memory. The desktop app reads how much is free and asks before opening a file that looks likely to need more, and you can open it anyway. A browser does not say how much memory is free, so there Olbo cannot warn first. On Windows, the desktop app keeps a file from a double click or a folder in memory once it has read it, and if a very large one will not fit there it says so; dragged onto the window, the same file is read straight from the disk.

Some limits no amount of memory moves, and they are the same on every plan. Chrome, Edge and the desktop app on Windows hold about 512 MB of text at once, so a CSV, JSON or text file larger than that does not open in them, and no file over about 2 GB opens in them at all. In Safari, and in the desktop app on a Mac or Linux, those two limits are about 2 GB and 4 GB. Olbo reads Excel and Word files of up to 4 GB, the most the zip format it reads can describe.

In the desktop app on Linux, and in other browsers built on WebKit there, an Excel workbook that unpacks to more than 448 MB does not open, whatever the plan. With that much of a workbook in memory, WebKit on Linux can stop responding for good, soon after opening it or while saving it. Chrome and Edge on the same computer open it.

Docs is limited by what a document holds rather than its size on disk. It opens up to 110,000 paragraphs in Chrome, Edge and the Windows app, 55,000 in other browsers and in the app on a Mac or Linux, and 12,000 on a phone or tablet.