I think we're going to see a lot more of this going forward.
I think we're also going to see the strategy to be to remove the processing and magic sauce from the client and move it to the server where it can't be decompiled and rebuilt with AI.
I think we're also going to see the strategy to be to remove the processing and magic sauce from the client and move it to the server where it can't be decompiled and rebuilt with AI.
I'd wager we will start to see more web apps like this have greater obfuscation and dependencies on operating on a particular domain. Sure AI can help to circumvent many things, but at a certain point they pay-off may not be worth the effort.
Also if it could recreate it that would be fine, because it would be doing so without having access to the source.
This doesn't sound that hard to automate these days.
I find it unlikely that photopea was never scraped for AI training considering they are looking so hard for new material they started buying up and scanning old books.
The magic sauce haven't been in the client for many applications for years. Google barely has any application that runs on desktop OS even though they could have released them.
It's a 1-man operation so it may have not have been financially viable to architect the app as server-based.
- server-based : must invest a lot more money in server farms and extra disk storage, or pay high AWS cloud fees. E.g. if a million users do a blur or denoise filter, all that cpu processing has to happen on the servers, and massive disk space to hold the intermediate files, and extra bandwidth costs to send the changed bytes back to the client.
- client-based : just ship Javascript blobs to end users' web browsers because the blur/denoise/etc filters happen on the desktop.
Also, this type of pixel-editing software still needs a ton of client-side Javascript to behave like a Photoshop clone because users want to see interactive changes as they dynamically slide the blur/noise/etc settings. Round-tripping that with extra server latency is not a fluid UI experience.
We can't confidently replay the past and say that starting it as server-side app from the very beginning means he'd have the same $million in revenue today. Instead, the extra server costs and UI jankiness could have doomed the project.
It's a business decision, and I'm not sure if they made the right decision. Most client-only web applications are open source because they know there is no business in selling it as a service. Photopea somehow is an exception, but its business model is getting questionable which is not a surprise at all.
It's also possible to use a combination of cloud based and local computation. Figma is doing quite well in that regard, especially with the use of WebAssembly. No doubt that potentially means more work, and potentially forcing users to create account etc. But hey, that's a business decision as well. If you don't do anything but just keep everything in JavaScript, this was going to happen.
Server-based photoshop clone sounds more like VNC/RDP, for this sort of thing client processing is a better UX.
SaaS killed Open Source with it, two decades ago.