Jan 15, 2026 Jan 14, 2026 2026-01-15

No, the title is not a mistake, it’s a reality. You can literally insert any app name there and it will still hold true. I know, I know, not quite ALL apps are replaceable by Claude Cowork, but still, a very sizable majority.

What Is Claude Cowork?

Think Claude Code, but for everyday tasks. If you are a coder (or a vibe coder), you already know what Claude Code is: the de facto AI tool for writing software, and it’s a damn good one. I’ve been using it for a few months already, as a developer (and not only) and I’m very pleased with it. It really makes my tedious tasks a thing of the past, and I can focus on high level architecture, bug fixing or adding features.

Now, Claude Cowork does the same, only not for the code. I know it’s a bit difficult to wrap your head around this.

So I’ll give you a few examples:

• you can organize some files on your computer
• you can ask Cowork to send messages (emails) for you
• you can create files
• you can crunch data from existing files and generate charts and diagrams

Claude Cowork is in research preview at the moment of writing, only available to Max users – but I honestly think this product was launched with market fit already.

The New UI Is Natural Language

I’ve been using an app called CleanMyMac for many years. It essentially scans my hard drive every once in a while and helps me get rid of the clutter. Identify huge files, leftovers, duplicates, and delete them.

I think you already know where I’m heading. Here’s a prompt I just used with Claude Cowork:

evaluate my Desktop folder and suggest improvements of the file organization. Some of them I still need, but it’s difficult to find them. The first thing that comes to my mind is organizing everything by year folders (maybe months inside year folders too?), but also some thematic structuring will be useful. Just give me your feedback, don’t do anything yet

It took Cowork about 5-6 minutes to:

• identify duplicates and delete them
• understand the type of file and its content (not only size or date, which CleanMyMac also does)
• create a semantically correct folder structure: Boarding Passes, Projects, Data Exports, etc
• move all the files around and show me the new structure

I find this impressive. And I think this hints at a completely new way (I was about to use the word “paradigm”, but let’s stick to “way” for now) in which we are using computers.

Before, we had visual interfaces with fixed layouts and actionable surfaces – buttons, checkboxes, menus. We were the ones initiating a workflow through these actionable surfaces, to generate some outcome.

Now, we instruct someone else about the outcome and things get done. That’s it.

But it goes even further. It can accomplish complex flows, involving several tools, for which there is no app yet. Read that again.

Here’s another prompt:

I want you to look in the Desktop folder and find me appTaskManager screenshots for Assess, Decide, Do and search functionality. I also want to use these screenshots to create a hero image 1256×640, with Assess, Decide, Do screens showing up the ADD framework.

Claude Cowork identified the screenshots, created the hero image with all the required constraints, here’s a part of its output:

The hero image is 1256×640 pixels and displays all three ADD framework screens (Assess, Decide, Do) side by side with color-coded labels matching your app’s theme (red for Assess, orange for Decide, green for Do).

I followed up with this prompt:

convert the hero image to .webp, make a folder called app_assets and move there the generated hero image, the containing iPhone screens as separate files, also .webp. and the hero search image, as separated .webp file

It did this in a few seconds. I estimate this workflow would have taken me maybe 10-15 minutes, on a good day. Cowork did it in less than a minute.

Endless Effectiveness

I think AI tools, and especially Claude Cowork – which seems to have found its market fit from day one – are becoming extremely effective now. I didn’t use the words “good at what they do”, because that’s not the point. They are very, very effective tools.

Imagine now that instead of prompting, we can chain a couple of other AI tools, like real time voice transcription and text-to-voice transform. That means we can actually talk to the machine. No more apps, no more UIs. Just endless effectiveness.

Pitfalls? Yes, Quite A Lot

While I find Claude Cowork extremely impressive, I think there are also some serious downsides. Some behavioral, some purely economical.

From an economical point of view, an entire app ecosystem will crumble. Maybe not today, maybe not next week, but we will see this unfolding before our eyes in less than 6 months. Apps will fold. Companies will close. Developers will switch jobs.

At the behavioral level, I already touched on this in a couple of posts here. If AI brings instant gratification, a.k.a. getting what we want instantly, then patience will become obsolete. If the friction involved in learning something new is gone, then we will literally become more stupid.

And last, but not least, if content production will become that easy, a lot of people will jump to the low hanging fruit of letting AI do everything, flooding the market with cheap, bad, but instantly available content. Because of this, I strongly believe bio content, or content generated by humans, will become a delicacy, carrying a significant premium.

Like this article, for instance. Not a word here was written with AI, yet I’m sharing my personal, live experience of using AI – which, in this current context, is like selling shovels instead of digging for gold.