Starting in April 2025, I went twelve months without installing a single new piece of software on my work laptop. Not as a stunt. As a working experiment. The constraint: anything I needed had to either already be on the machine or run in a browser tab. I could update existing apps. I could not download new ones, including command-line tools that lived outside my package manager. The point was not to be pure about it. The point was to find out what the install button is still actually for.
A year in, I have a clear answer. The install button is for three things: media, code, and games. Almost everything else I used to install has a tab now, and the tab is usually better.
What I expected to miss, and didn't
Office software was the first surprise. I had assumed I would crawl back to a desktop editor within a month. I did not. Google Docs is fine. Notion is more than fine. The collaboration features that show up on the desktop versions of Office and Pages are actually worse than the browser equivalents now — Microsoft's strategy team apparently decided this two years ago and stopped polishing the standalone apps.
Design was the second surprise. I had been a holdout on Sketch through 2023 and I expected to feel the loss. Figma covered everything I had been doing in Sketch and three things I had not realized I wanted (real version history, real-time co-edit, a dev-mode handoff that finally killed the 'export specs to PDF' ritual). For the once-a-month image work, Photopea handled it. I never opened Pixelmator the entire year.
Email I had already moved to a browser long ago. Calendars too. Slack was already a browser tab in disguise. So was Discord. So was, in practice, the music app, since I had switched to the Spotify web player in 2024 and never looked back.
What broke
The first failure was video editing. I tried. There are real, capable browser editors now — Veed, Clipchamp, even a credible Descript alternative — and for the kind of edits I do (a five-minute talking head, with some b-roll, a few captions), they were almost good enough. Almost. The export speed was 3-4x slower than the desktop equivalent, the timeline scrubbing got laggy past about six minutes of footage, and the keyboard shortcuts were a constant fight with the browser's own. I cheated twice in the year. Both times I regretted not cheating sooner.
The second failure was coding. I started with the assumption that GitHub Codespaces and StackBlitz would cover me. They covered about 80% of what I needed. The remaining 20% — anything involving local databases, anything requiring more than 8 GB of memory, anything where the latency to the editor mattered for flow — was painful. The web-based IDEs have improved every quarter for three years and they still aren't quite there. By June I had quietly grandfathered VS Code in (it was already installed) and stopped pretending.
The third failure was specialized tooling. There is no good browser-based alternative for the OBS I used to record screencasts. There is no real browser-based equivalent to ffmpeg for the times I need to convert one weird video format to another. There is no browser-based equivalent to the password manager browser-extension behavior that some workflows actually do require. These are small, specific holes. They are also genuine.
What I gained that I did not expect
The biggest surprise was setup time. I changed laptops in November — switched to a new machine in two hours, which is genuinely the fastest I have ever switched machines in my life. The browser logged in. Everything was there. No re-installs, no license keys, no 'where did I put the configuration file for this thing,' no afternoon spent re-downloading the same five applications and discovering that two of them now require a different paid tier than they did when I bought them last time. Just: open the browser, log in, work.
The second surprise was discipline around state. When all your tools live in a browser, you stop accumulating private state on the machine itself. I deleted nothing important when I migrated. There was nothing important to delete. Everything that mattered was already on a server somewhere, owned by an account I controlled, exportable on request. That was an unfamiliar feeling for someone who has been carrying around the same /Documents folder, with slowly accumulating fossil files, since 2012.
The third surprise was that I started caring more about what I let into my workflow. The install button is fast. It is so fast that you do not really evaluate the program — you just install it, decide it sucks, and forget to uninstall. The tab button is also fast. But each new tab is one more bookmark, one more thing that wants your attention, and the lightness of the medium makes you ruthless about what stays.
The honest counterargument
Browser apps are dependent in a way that desktop apps are not. When my internet went out for forty minutes in February — a router issue, not a heroic outage — my whole workflow stopped. The browser apps I trusted to have offline support mostly did. The ones I assumed had offline support mostly did not. PWAs in 2026 are real but uneven. Some apps will let you keep editing for hours unplugged; others throw up a stale spinner the moment the connection drops. You learn which is which only after it matters.
There is also a category of work that benefits from a heavyweight, opinionated, native interface — DAWs, video editors, 3D tools, the genuinely demanding creative tools. The browser is closing the gap. It will probably close most of it. It has not closed it yet, and pretending otherwise would be embarrassing.
Where this is going
Three forces are pushing the install button toward obsolescence for general-purpose work. First, the browser has become genuinely capable — WebGPU, WebAssembly, the File System Access API, the persistent storage permissions are all just there now, and a determined team can ship a serious application without leaving the tab. Second, IT departments hate native apps. The friction of provisioning, patching, and securing endpoint software has gotten high enough that 'this runs in a browser' is now a category that wins enterprise deals on its own. Third, users have noticed that the install button is mostly used against them — as a place to put updaters that never turn off, telemetry that does not ask, and renewal pop-ups for software they stopped using two years ago.
I do not think native software is going away. I think it is going to retreat to the workloads it is genuinely better at, which is a smaller list than the desktop industry would prefer to admit. Everything else — the long tail of utility software that defined the 2000s — will live in tabs.
If you want to try it
You do not have to do a whole year. A reasonable version of this experiment is: for the next sixty days, do not install anything new. When you reach for the install button, look for the browser equivalent first. If the browser version is 80% of what you need, that is usually enough to learn what you actually use. If it is not enough, you will find out fast, and you will learn something useful about which class of work still needs the install. The exercise is mostly about discovering how much of your workflow is already a tab in disguise.
I will probably not install another general-purpose app on this machine voluntarily. I might lose this argument in 2027. I do not think I will.
The catalog on dotstore is, mostly, my year's worth of replacements organized by category. If you are trying the sixty-day version, the Productivity and Developer Tools sections are the two I would start with.
Tags