There is a small software product most people walk past every day, and the reason it has survived for over a decade is the same reason nobody has ever tried to file a feature request for it. The team that built it picked a design choice that would get a normal product manager fired in any other context. They built something destructive, they refused to add an undo animation, and they trusted the user to hit the reset button themselves. That choice is the whole product. It is also a quiet argument about how software should handle irreversible actions, and it is one of the better ones I have seen in the last twenty years.
The trick itself is a fan. You type a phrase into the search bar, a giant fan appears in the middle of the page, and the layout starts blowing apart. The letters do not shake. They tear off one at a time and drift toward the right edge of the screen until they are gone. Crank the speed up and the whole page leaves in one motion. You end up staring at an empty white screen with a fan still spinning at nothing, because the page that used to be there is no longer there. That is the joke. It is also the whole point, and the engineering choice behind it is worth talking about.
The reset button is the only thing between chaos and a clean page
Once the page is empty, the fan does not slow down. It does not reverse direction and try to put the letters back. It does not slowly rebuild the layout over a few seconds. It just keeps spinning at a target that no longer exists, which is the part most modern product teams would find uncomfortable. The instinct in 2026 is to undo state changes automatically, especially ones that look destructive. If a user accidentally deletes something, the system restores it before they finish their next click. If a button does something surprising, the UI reverses it after a short delay. The default assumption is that any visible state change is a bug, and the job of the product is to make the bug invisible.
The fan does the opposite. It commits to the destruction and trusts the user to recognize that they want it back. The reset button is one click away, the layout snaps back to normal the moment you press it, and there is no penalty for triggering the fan as many times as you want. The team that built this could have added a three-second auto-rebuild. They did not, because the moment where the page is empty and the fan is still spinning is the moment that makes the trick funny. Auto-rebuild would turn the trick into a cartoon. Without it, the trick stays absurd, and absurd is harder to build than cute.
I think that is the right call, and I think it is a lesson most modern product teams have stopped believing in. The fan trusts the user to know what they clicked. It trusts the user to know that the button is still there. It trusts the user to press it. That trust is a feature, and it is a feature that costs almost nothing to build, because the only thing you have to do is leave the destruction alone.
Five speed settings are a small act of restraint
The fan has five speeds, and the difference between them is the entire product. Low wobbles the letters in place. Medium pulls them off the page. High shreds the whole layout. The top setting sends everything off the screen at once and leaves the fan spinning in a void. The interesting engineering choice is not the speed settings themselves. It is that there are exactly five of them, and the team never added a sixth.
You can imagine what a sixth speed would look like. A mode that auto-feeds the fan new material so the page keeps getting destroyed. A mode that adds a second fan at the bottom of the screen. A mode that changes the sound. None of those things exist, and the reason they do not exist is that the team built a trick instead of a product. A trick has one good version. A product has a roadmap. The fan is a trick, and the team has been disciplined enough to leave it that way for the entire lifespan of the feature.
The five speeds are also a small example of how good interface design works. You do not need a slider. You do not need a settings panel. You do not need a tooltip explaining what each speed does. You click the speed you want, the page does the thing, and you can see the difference between the speeds the first time you try them. The control surface is small on purpose, because the point of the fan is to be triggered, not to be configured.
Irreversible actions are a design choice, not a constraint
The broader pattern here is that most software in 2026 treats irreversible actions as a problem to be solved. If a user deletes a file, the system asks them three times if they meant it. If a user closes a tab, the system offers to reopen the tab group. If a user cancels a subscription, the system asks them to confirm, then asks them to take a survey, then sends three emails over the next two weeks offering to bring the subscription back. The whole industry has decided that any action a user might regret is an action the system should slow down, confirm, or undo automatically.
The fan is one of the few products on the internet that does the opposite. It does something the user asked for, it lets the user see the full effect of what they asked for, and it offers a single button to undo the thing. That is the whole interaction model. No confirmation modal. No undo toast that disappears after three seconds. No “are you sure you want to destroy this page” prompt. Just the action, the consequence, and the reset button.
I think the fan works because the consequence is visual and reversible, which is a rare combination. Most irreversible actions in real software are also invisible to the user. They delete data. They submit a form. They publish a post. None of those have a visible effect the user can watch before the action completes. The fan is unusual because the destruction happens slowly enough for the user to see it happening and fast enough for the joke to land. That is the design constraint the team is exploiting, and they are exploiting it well.
The fan has no sequel and that is the whole point
People search for “google fan 2” and “google fan 3d” constantly, looking for an advanced version that does not exist. There is also no “google fan game” in any app store, no extension that adds a sixth speed, no companion site that hosts a longer version. The fan is one page, five speeds, and a reset button. That is the entire product.
The scarcity is the joke. If there were a sequel, the trick would stop being a discovery and start being a franchise. People would compare the two versions, write hot takes about which one was better, and argue about whether the sequel ruined the original. None of that happens, because there is nothing to compare. The fan is the only fan, and the team has kept it that way for over a decade.
This is also why the fan has survived. It is small enough to ignore, weird enough to remember, and good enough to recommend. The team did not need a roadmap. They did not need a content calendar. They did not need to update the fan every quarter to keep it fresh. They just left it alone, and the thing kept working.
Trade-offs
The fan is a small product, and most of the trade-offs are about what is missing.
- No auto-rebuild: some users will trigger the fan and then wait for the page to come back on its own, which it never does. That is the point of the joke, but it is also a real friction point for people who do not know there is a reset button.
- No mobile version: the trick does not work on phones, because the desktop browser is where the search bar still behaves like a search bar. Anyone trying to share the trick on mobile will be confused.
- No settings panel: there is no way to save a preferred speed, no way to swap the fan for a different visual, no way to slow down the destruction. The control surface is exactly what the team shipped ten years ago.
- No companion API: you cannot embed the fan in your own page, hook it into a script, or trigger it from a URL parameter. The trick only works on the actual Google search page, which is what makes it feel like a discovery instead of a feature.
What I would tell past me
The fan is the kind of product most engineers wish they could ship. Small enough to fit in a week of work, weird enough to be memorable, and good enough to survive a decade without anyone trying to “improve” it. The lesson is not about fans. It is about the discipline of leaving a finished thing alone once it is finished, which is a discipline most of the software industry has forgotten.
If you have not tried the trick yet, open a desktop browser and type the phrase into the search bar. Crank it to the top speed once. Watch the page disappear. Press the reset button. That is the whole product, and the whole product is the point.