
last week, anthropic launched claude cowork, inside the claude desktop app. at a high level, it allows users to:
point claude at a folder on your computer
ask claude to do real work (read files, summarize them, rename them, reformat them)
from there, claude will actually make those changes, directly in the folder
that’s the key shift from before:
previously: you uploaded files to claude, asked for help, and got answers back in a chat window
now: you give it folder access, and it acts directly on those files for you
tbh, the current use cases feel only moderately useful: summarizing call transcripts, cleaning up messy filenames, pulling receipts into an expense report. nothing life changing.

product mock from anthropic’s launch materials
however, i do like that instead of yet another agentic feature targeting developers, claude is helping normal people get through annoying, time-consuming computer chores.
that said, i am skeptical about adoption. most people still don’t know what claude is, let alone have the desktop app installed.
imo, the real opportunity here is for macos and windows to steal this idea. imagine this as a native right-click option, next to the actions you already use every day. apple and microsoft could ship this to hundreds of millions of users overnight. no app to download. no new workflow to learn.

quick macos mockup from figma make
as i’ve written before, i believe operating systems have an extreme right-to-win for many ai features. this is a great example of that.
in summary: claude co-work feels like a solid, truly “agentic” feature that signals where the future of work is going. i just don’t think this current incarnation will see much traction. that moment will arrive when windows or macos completely steal this idea builds this themselves.
