Codegres.org VS SafeUnfollow

SafeUnfollow's answer:

SafeUnfollow never asks for your Instagram password, and it has no server to send anything to.

It works from the data export Instagram already lets you download - the same ZIP Meta emails you when you request "Followers and following". You drop that ZIP on the page and the whole comparison happens inside the browser tab: the archive is unpacked in a Web Worker and the lists are held in the browser's own IndexedDB, stored column-wise with bitset filters so it stays responsive even on very large accounts.

Two things follow from that design, and they are the point of the product:

• There is no upload step, so your follower list never reaches us, a processor, or any third party. We cannot leak what we never receive.
• There is no login and no call to Instagram's API, so using it involves no credential and no automation against your account.

It keeps working offline after the first load, and it would keep working if the site disappeared tomorrow. The analysis is free at any account size and the code is MIT-licensed on GitHub.

SafeUnfollow's answer:

Choose it if any of these three things matter to you.

You do not want to hand over your Instagram password. SafeUnfollow has no login screen at all. It reads the export Instagram itself gives you, which is the route Meta documents and supports, so there is no credential to share and nothing signing in on your behalf.

You do not want your follower list sitting on someone's server. There is no upload step and no backend. The ZIP is parsed in your browser and stays in your browser's storage, which you can clear whenever you like.

You do not want a subscription. The full analysis is free at any account size - every badge, every filter, however many accounts are involved. The only paid item is an optional $7 one-time licence to export the results table, and a free 10-row sample of that export is shown before you decide.

Two more that matter to some people: the code is MIT-licensed and public on GitHub, so the privacy claim can be audited rather than trusted; and the interface runs in ten languages, including Indonesian, Spanish, Portuguese, Russian and Arabic.

What it is not: SafeUnfollow does not unfollow anyone for you, and it is not a real-time tracker. It reads a snapshot you download, so comparing over time means requesting a fresh export.

SafeUnfollow's answer:

Three groups, and they overlap.

People who are wary of unfollower apps. The category has a reputation for asking users to sign in with their Instagram credentials, and many people arrive having decided they will not do that. They search in terms like "without login" and "with the ZIP file". SafeUnfollow is built for that reader first: there is no login screen to reach.

People with large accounts. Creators and community managers whose following list is long enough that lighter tools stall or cap out. The storage layer is columnar with bitset filters precisely so the list stays usable at that size, and there is no account-size tier or row limit.

People who will not pay a subscription for this. Checking who unfollowed you is an occasional curiosity rather than a monthly need. The analysis is free in full, so it fits that shape.

The audience is not English-only, which is why the interface ships in ten languages, including Indonesian, Spanish, Portuguese, Russian and Arabic.

SafeUnfollow's answer:

TypeScript throughout, React 18 with Vite, and Tailwind CSS with shadcn/ui (Radix primitives) for the interface.

The interesting parts are the storage and compute layers, because everything runs on the user's own device:

• JSZip unpacks the Instagram export inside a Web Worker (wired up with Comlink), so a large archive never blocks the interface.
• IndexedDB, used directly rather than through a wrapper, holds the data column-wise: usernames and profile links as packed byte arrays, and each badge as a bitset - one bit per account - instead of a boolean array.
• FastBitSet does the filtering as bitwise set operations over those bitsets, so combining filters is set arithmetic rather than a scan.
• TanStack Virtual renders only the rows actually on screen, which is what keeps a very long list scrollable.
• Trigram and prefix indexes back the username search.

The public pages are pre-rendered at build time with vite-react-ssg, and the app is an installable PWA (vite-plugin-pwa) that works offline after the first load. Localisation is i18next across ten languages. The documentation site is Jekyll on GitHub Pages; hosting is Vercel.

There is deliberately no backend, no database of ours and no API of our own. The reason the export never leaves the device is that there is nowhere for it to go.

SafeUnfollow's answer:

It starts from a simple observation: Instagram already gives you the answer. Meta's own data export contains your followers list and your following list, and "who unfollowed me" is a set difference between two files you can download for free. The information does not have to be fetched from anywhere.

Any design that skips the export has to obtain the same lists some other way - by signing in as you to read them, or by taking a copy onto a server. SafeUnfollow was built to need neither. Take the ZIP Instagram gives you and do the comparison where the file already is: in the browser tab.

That one decision removes most of the surface a tool like this would otherwise have. There is no account system, no server-side history, no scheduled scanning, and no way to unfollow anyone from the app, because none of it is possible without the things it deliberately does not have. What is left is the part people actually came for, and it is free at any account size.

The code is MIT-licensed and public for the same reason. A privacy claim you cannot check is a marketing sentence; one whose source you can read is a property of the software.