IDK, but I have three or four vibe-coded 0.1%-Photoshops already - small utility apps, all self-contained, client-side, static, zero build-step browser tools, each of them solving a very particular problem for me.

Things like, drop an image, get a printable label sheet that force-overrides all margin and scaling settings (recent upgrade to a tool I had that did that with plaintext address data), or metadata-preserving image cropper with some OpenCV magic to auto-straighten images and then allow free-form cropping that snaps to lines and features in the image, etc.

All of those have three things in common: they're dedicated to a very specific task or class of tasks that I happen to be doing, they each took less than an hour to create, and they exist specifically so I don't have to deal with whatever incidental bullshit that Photoshop or Word or Affinity or other Professional Tools For This Job throw at me.

(The label printer thing in particular was a pure frustration job.)

Pretty sure there's tons of such vibe-coded apps around - that do their jobs well, displace the classical general-purpose tools for their authors, but also they're too specialized to try and productionize/release.

EDIT: more recently, I even gave up on creating some of the apps in cases I normally would, after I noticed I can talk Claude on the mobile app into writing and running Python code, which turns out sufficient for many tasks involving data and image analysis.

Yes, this is the point.

LLMs are approaching the affordances that OpenSource claimed to offer but never did.

Non-coders are discovering the basics of - for example - Python scripting and running with it.

If this gets easy enough - questionable so far, but maybe one day - all code will be "Do this thing to this photo/document/music project."

The big toolbox products will become redundant.

Everyone is distacted by content generation, but if LLMs get good enough it will be possible to go back a step and give everyone a toolbox factory they can use to imagine and build whatever they want, with full control, instead of the current stab-around-until-you-find-something generative approach.

Exactly what will eat into PhotoShop etc.

I suspect that if you start seeing vibecoded "PhotoShop alternatives" productised, they won't look like PhotoShop at all: They'll be something like a bunch of image manipulation tools with an agent loop and scripting, to make it easier to create and run a bunch of 0.1%-Photoshops on-the-fly and save/reuse them.

So something more like Claude's artefacts support, just hiding the techie bits for users not comfortable with that, with specialised tooling.

I'm guessing we'll see a bunch of apps like that eating into various traditional big monolithic tools.

And most importantly, you don't care at all if the tool you vibe-coded is any good. If you write at tool that converts an image to black & white, you are the kind of person who doesn't know or care what KIND of black&white it is. The fact that there are many algorithms to choose from would never cross your mind.

The same happens with whatever tool you vibe-code. You get the average of the worst quality open source versions that exist, combined with some randomness.

This is not "do one thing and do it well", this is "do one thing no matter how".

It's almost like people are finding out what scripting is because ai can do it for them.

This isn't new most businesses are run on scripts that process data. The only difference now is that more people can write them instead of paying for an app.

So the narrative here is wrong. Because vibe coding an app is overkill, in the same way that paying through the nose for a giant app that you only use one basic feature set of is.

[flagged]