in the first hour of Metal Gear Solid 2, one of the guards is surrounded by flies.
he’s carrying a ration that has gone bad. if you take the ration from him, the flies come with you. And because they follow you, they will give away your position to the next guard down the corridor.
You just learned that a piece of set dressing turns out to be a rule, and the rule has consequences.
shoot a soldier in the arm and he holds it, moves differently (like like a man with one working hand). or enter in room when you’re wet and you’ll leave footprints behind you that can alarm guards.
and you can’t even imagine all the details put in this game.1
some details are so insane that Kojima had to reveal them himself, on Twitter, years after the release.2
I am obsessed with this kind of work, and for a long time I could not explain why.
Is it just an “attention” to details like people say?
Prove that you care
Anyone can claim care. Marketing claims it. Landing pages claim it. How much “Crafted with love” i saw on a footer without real evidence.
these claims are free, it absolutely prove nothing.
Details cost: hours spent on small details cannot be faked.
This is why Sam Esmail’s technical obsession on Mr. Robot is not fan service.
Almost each website, IP address, phone numbers displayed on the screen was real.
https://www.youtube.com/watch?v=5dLRNoS8H-A
i think the same attention must be applied to software.
Modern products are optimised toward the median path. Features are justified by usage data, tested toward the mean, and cut when adoption is low.
If one user in a thousand would ever open the fridge, the fridge does not get built, because one in a thousand is indistinguishable from noise on a dashboard.
The curious user is a rounding error, and rounding errors get designed out.
It is worse than indifference. Attention-optimised software actively punishes curiosity, because exploring means leaving the funnel. There is nothing to poke at, since anything worth poking at would delay the next impression. The feed does not want you inquisitive. It wants you receptive. So we ended up with an entire industry that treats the most valuable instinct a user has as a leak in the conversion rate.
The new way to lose it
in my previous note, i have written that ambient AI is the right direction, but how it will impact the user curiosity?
I think the intelligence layer in the next software must be able to open something nobody promised was openable. A system that anticipates you perfectly does not lock that door, it hands you what was behind it before you thought to walk over.
A model trained on everything produces the middle of everything. It is astonishingly good at the expected and structurally uninterested in the unnecessary. No system optimised on a reward signal will ever decide, unprompted, to simulate ice melting in a room the player has no reason to enter. Nothing in the objective function pays for that. Nobody asked for it. That is the entire point of it.
Reward the curiosity
Reward it. Everywhere, in every context you control.
Reward it in your users : leave something behind the door they were not required to open.
Reward it in the people you work with : the question asked outside the scope of the ticket is the only reason anyone ever finds the real bug.
Reward it in yourself, especially now, when the answer arrives before the question has finished forming and wandering has never been less efficient.
Build the thing nobody will ever see. Some of them will find it anyway, ten years later, and they will understand immediately what it meant. And the ones who never find it will still feel it. That is the strange part. They will not be able to say why the thing felt solid under their hands. But they will know.
if you care about details
i recommend you to watch these videos:
Footnotes
-
This game was released in 2001. There’s a full rundown if you’re curious. ↩
-
For example, the ice in the bar melts — and the cubes packed together melt more slowly than the ones sitting alone. ↩