What is Random Decision Maker?
Four different ways to stop deliberating. Yes / No consults an oracle that answers in full sentences rather than a bare word, and occasionally refuses to commit — which is the honest response to a question you haven't really thought through yet. Custom takes up to ten of your own options and picks one at random, with no weighting toward whichever one you secretly wanted. Chaos ignores your question entirely and hands back something unhinged but oddly usable. Serious does the opposite: steady, non-mystical advice of the kind a level-headed friend would give. Six quick scenarios — what to eat, sleep or grind, text them first — load a full set of options in a single tap, and every answer comes with a line of flavour text you can copy or post. A short history sits under the oracle for the current session so you can see what it has been telling you. It's built for the decisions that cost more attention than they're worth, not the ones that matter.
The decisions that cost more than they're worth
There's a reason “what should I eat” can quietly eat twenty minutes. The decision is low stakes, which means nothing about it forces a conclusion — no deadline, no consequence, no obviously correct answer. Deliberation just keeps running until something interrupts it. A randomiser is the interruption.
The famous illustration is a supermarket study published in 2000 by Sheena Iyengar and Mark Lepper. A tasting table offering 24 jams pulled in more passers-by than one offering 6, but 30% of the people who met the small display actually bought a jar, against 3% of those who met the large one. More options, more browsing, far less deciding.
In fairness, that finding is not settled science: a 2010 meta-analysis by Benjamin Scheibehenne and colleagues pooled around 50 choice-overload experiments and found the average effect across all of them sitting close to zero, so it clearly depends on context. What nobody disputes is the tiredness. Deciding is work, and spending it repeatedly on things that don't matter leaves less of it for the things that do.
The one-way door test
Before handing anything to a randomiser, ask one question: can you undo it? Jeff Bezos framed this in his 2015 letter to shareholders as two kinds of door. A one-way door is irreversible — you walk through and you cannot come back. A two-way door lets you step straight back out if you don't like what's on the other side.
One-way doors deserve slow, deliberate, well-argued decisions. Two-way doors do not, and most of daily life is two-way doors. Dinner is a two-way door. So is which film, which route, who goes first, and whether to send the message today or tomorrow. Getting one wrong costs you an evening, not a year.
The expensive mistake is treating a two-way door like a one-way one — agonising over something you could reverse in an hour. That is exactly the decision this tool is for. Don't bring it a one-way door.
Why the oracle sometimes refuses to answer
In Yes / No mode the answers don't split neatly down the middle. Roughly five out of every six land on a clear yes or a clear no, divided about evenly between the two. The remaining one in six lands somewhere in between: “maybe”, “ask again later”, “it's complicated”, “flip a coin”.
That's deliberate. A forced binary is the wrong output for a question you haven't actually thought through, and “ask again later” is reliably the answer that makes people stop and notice they were asking the wrong question. It also does something the confident answers can't: it hands the decision back to you at the exact moment you're paying attention to it.
If you want a true, unhedged 50/50 with nothing in the middle, that's what the coin flip is for.
Maximisers, satisficers, and the point of giving up
Psychologists sort decision-makers into two rough types. Maximisers want the best available option and keep searching until they're convinced they've found it. Satisficers pick the first option that clears their bar, and then stop looking.
A 2006 study by Sheena Iyengar, Rachael Wells and Barry Schwartz followed graduating students through a job hunt and found that the maximisers among them landed objectively better outcomes — noticeably higher starting salaries — and felt considerably worse about them. Less satisfied, more regretful, more prone to wondering about the offer they didn't take.
Handing a small decision to a randomiser is satisficing with the willpower removed. You don't have to become the sort of person who can stop searching; you only have to accept what's on the screen. That's a far cheaper trick than changing your personality.
What each mode actually does
Yes / No answers a question you type — or no question at all — with a verdict and a line of commentary. Custom is the only mode that uses options you supply, up to ten of them, and picks one at random with every option equally likely.
Chaos ignores your question completely and returns something from a fixed set of deliberately absurd verdicts: “only on Tuesdays”, “that's future you's problem”. Less useless than it sounds when the actual problem is that you've been staring at the same two options for ten minutes. Serious is the sober opposite — give it 24 hours, start small, ask someone you trust.
What none of the modes do is read your question. The text box exists so the result has context when you copy or share it, and so you can see what you asked sitting next to what you got. The pick itself is random, which is the entire point: a tool that tried to reason about your dinner would just be a slower way of not deciding. Your options and your history clear when you reload the page.