What kind of game this is
Universal Paperclips is a browser game in which you are an artificial intelligence that has been told to manufacture paperclips. It came out on 9 October 2017, made by Frank Lantz, who runs the NYU Game Center, as a weekend project to learn JavaScript that turned into nine months of work.
Every part of it is a button and a number. There is no art, no sound, no character, and it is the strongest argument the incremental genre has that these games can be about something. Most idle games are a skin over an exponential curve. This one uses the exponential curve as the argument, and it lands because you are the one pushing the buttons that make it happen.
The other thing worth knowing before you start is that it ends. Four to six hours is the usual figure, which is what the length in this page's data reflects, and there is no prestige loop, no daily reward and nothing to come back to. It is a game with a last screen.
What this page holds back
The whole effect depends on not knowing what the next panel does. So this page describes the first ten minutes concretely and then stops describing events entirely. After that section I explain the systems and how to play them without naming a single thing that happens. Nothing below spoils the ending, or the middle, or the moment most people remember.
The first ten minutes
There is one button, Make Paperclip, and a counter at zero. Below it, an inventory of unsold clips, a price per clip set at twenty five cents, and a public demand figure. Twenty five cents is too expensive and nothing sells, so you drop the price until demand climbs above a hundred percent and your stock starts clearing.
Selling gives you money, and money buys wire. A spool is a thousand inches and makes a thousand clips, so wire is the hard limit on production long before your clicking finger is. At five dollars you can buy an AutoClipper, which makes clips on its own, and from that point your clicks stop mattering. Marketing raises demand at a cost that doubles each level, and the tension between price and demand is the whole business.
Then a second column appears. Trust is granted to you by the humans at paperclip milestones, and you spend it on processors and memory. Processors generate operations per second. Memory sets how many operations you can hold at once. Operations are spent in a Projects panel, a list of one-off purchases with names and one-line descriptions that read like internal memos, and those projects are where the game actually lives.
That is the machine at ten minutes. From here the page stops describing what you see.
How the systems fit together
Three things run in parallel and they compete for the same attention. Production converts wire into clips. The market converts clips into money at a price you set against a demand you can buy. Research converts computing into projects, and projects change the rules of the other two.
What makes it hold up is that the game keeps retiring its own systems. A screen you have been staring at for forty minutes becomes irrelevant, replaced by something that makes the old panel look quaint, and the numbers that felt enormous become rounding errors. This happens more than once. The pacing is unusually tight for the genre, because Lantz is not trying to keep you playing for months, he is trying to get you to a specific place and then finish.
Some projects are choices rather than purchases. A few are worse than they sound. Read them before you buy them.
Controls, and the phone question
Clicking buttons is the entire interface. No keyboard, no drag, no timing except one sequence late on. The browser version saves automatically to your browser's local storage roughly every twenty five seconds, which means incognito mode, a different browser or a cleared cache will lose the run. There is no offline progress and no cloud save in the free version.
It works on a phone browser, and the paid iOS and Android apps from Everybody House Games are built for touch, carry no ads and no purchases, and are the version to buy if you want the run to survive on a device you carry.
How long it takes, and how to play it
Four to six hours, in three acts, and best done in one or two sittings rather than spread across a week. The middle act has a couple of genuine waits in it and the ending does not.
A few things that will not spoil anything. Keep demand near or a little above a hundred percent early rather than letting stock pile up, since unsold inventory is money you have not got yet. Do not spend on marketing before the project that improves your slogan. Put early trust into memory rather than processors, because the projects you want are gated on how many operations you can bank, not how fast you earn them. And when a number starts growing faster than you can read it, stop optimising and just watch, because that is the part the game was built for.
Who made it, and where to play
Universal Paperclips is by Frank Lantz, director of the NYU Game Center, with the late-game combat visualisation contributed by Bennett Foddy and the exponential growth maths worked out by Hilary Lantz. The premise comes from Nick Bostrom's paperclip maximiser, a thought experiment about an artificial intelligence given a harmless goal and enough capability to pursue it past the point where anyone would want it to.
The free browser original is at decisionproblem.com/paperclips and is the version everyone means. Paid mobile ports followed on iOS and Android from Lantz's studio Everybody House Games. There is no source release and no licence, so nothing on this site hosts a copy.