Siloes

If I Ran Silo 1: How I’d Redesign the Algorithm

Spoiler warning: this post covers Silo up to and including the Season 3 finale, “Troy”.

For three seasons, the Algorithm has been the most powerful presence in Silo, an unseen system judging the stability of Silo 18 and deciding whether 10,000 people live or die. Season 3 finally pulled back the curtain. There is no machine. There is Victor Crnkovich in Silo 1, speaking through a voice modulator, with Rosalind Thurman as director and a memory-wiped Daniel Keene, now “Troy”, woken from cryosleep whenever a silo needs putting down.

The moral case against Silo 1 hardly needs making. What interests me is that, even on its own chilling terms, the system is badly designed. If the goal is to keep 50 isolated populations obedient for centuries, the founders made a string of mistakes any strategist should have spotted. So here is a thought experiment: if I were the Algorithm, what would I have done differently?

The core flaw: power built on a lie
Silo 1’s control rests on a shared false belief. The residents are not only told that the outside is lethal and that the Algorithm is infallible; they are kept from their own history altogether. The project’s stated aim is to free humanity from its past.

The trouble with ruling through a false belief is that it is a coordination problem waiting to be solved. Obedience depends less on what each resident privately believes than on what they think everyone else believes. Once the truth becomes common knowledge, when people know that others know, compliance can collapse almost overnight. Nearly every crisis in the show follows from Silo 1 creating ways for that knowledge to spread.

1. Don’t send dissenters to the one place they can test the lie
Cleaning is meant to be the ultimate public deterrent: challenge the system and you walk out to certain death while everyone watches. It works perfectly, right up until someone survives.

It is an astonishingly risky design. The punishment puts the most curious, rule-breaking people exactly where the fake display on the screens can be exposed, and one survivor, Juliette, was enough to shatter it. A deterrent with even a small chance of disproving your entire narrative in front of the whole population is not worth having. I would handle punishment inside the silo, out of sight of the outside world, or at the very least make sure no engineer in Mechanical could quietly defeat the suit.

2. Never give one person all the knowledge
In Silo 18, the head of IT is effectively the only person who knows the truth. That puts the whole regime at the mercy of one person’s temperament, ambitions and succession politics. Bernard’s paranoia, his clumsy choice of successor and his eventual turn against the system show how badly that can go.

A sturdier design spreads knowledge across several people who don’t fully trust each other and can check each other’s behaviour. Just as important, Silo 1 needs independent ways of watching what is going on. In the finale, Camille’s feint worked because Victor was relying on a single human channel he could not verify. An Algorithm that can be bluffed is not much of an Algorithm.

3. Make the threat automatic, or keep it secret
In game theory, a deterrent is most credible when the side issuing it cannot back down. Silo 1 got this the wrong way round. The Safeguard ran through a physical pipe that residents could find and cap, so they knew the threat was real and also that it could be beaten. Worse, it was run interactively by a person, which told everyone there was someone at the other end who could be negotiated with, delayed or tricked.

I would choose one of two coherent options. Either the threat is fully automatic, triggered by fixed and published conditions with no human discretion and no physical point of failure, or its existence is never revealed at all. The half-hidden, human-operated, breakable version Silo 1 used combines the weaknesses of both.

4. Punish in proportion
The Directive treats contamination between silos as the one unforgivable offence, and entire silos are expendable to prevent it. That is the equivalent of a strategy that punishes everyone for a single defection, and over repeated rounds it has two fatal problems.

First, it destroys the very thing the project exists to protect. With 50 silos and centuries to go, wiping out whole populations steadily drains the stock of humanity the founders claim to be saving. Second, it removes any reason for restraint. Once a silo believes it is already condemned, it has nothing left to lose, which is exactly the point at which people rise up. Graduated, targeted sanctions keep deterrence working while leaving people a reason to stop.

5. Take the mystique out of the past
The total ban on relics was meant to cut people off from history. In practice it made every surviving object enormously significant. A forbidden children’s book or a single photograph becomes proof that something was hidden, and hidden things invite investigation.

I would do the opposite: allow a carefully chosen supply of harmless old-world objects so they become ordinary. Curiosity also needs a safe outlet. The most inquisitive residents are the most dangerous if you fight them and the most useful if you recruit them. Bernard half understood this with Juliette, but by then trust had already broken down. A system that routinely brought its brightest dissenters into IT, rather than sending them out to clean, would face far fewer rebellions.

6. Don’t build what you’ve banned
If isolation between silos is the top priority, a network of tunnels connecting them is a baffling weakness. It is like announcing that doors are the gravest threat to the regime and then fitting every wall with one. Either don’t build them, or make their existence the most closely guarded fact in the entire project. By the end of Season 3, Juliette knows about them, and that is catastrophic for Silo 1.

7. Keep the enforcer reliable
Troy is Silo 1’s troubleshooter, the person trusted with the hardest decisions. He is also plainly deteriorating: each time he is woken, flashes of Helen come back and he asks more questions. Victor’s response is simply to put him back to sleep.

That is putting off the problem, not solving it. The person with the most lethal authority in the system should be its most predictable part, and Silo 1 is relying on memory suppression that visibly fails. Sooner or later, the enforcer remembers.

8. Make promises that can be checked
In the finale, Troy offers Silo 18 a deal: he will leave the Safeguard off as long as they avoid contact with the other silos. Juliette seems to accept, while planning to reach Silo 1 through the tunnels.

Of course she does. Silo 1 has lied to her people for generations and has shown it will kill entire populations. A promise from an actor with that record is worthless unless it can be verified or enforced. If I genuinely wanted a settlement, I would give her something concrete and lasting: real control over the capped pipe, independent monitoring, or even a seat in the project itself. Commitments only hold when breaking them costs the one who made them.

The uncomfortable conclusion
What strikes me about this list is that almost every change makes the system less cruel as well as more stable. That isn’t a coincidence. Regimes held together by deception and collective punishment are fragile precisely because they give people every reason to resist and nothing to lose by doing so. Legitimacy is far cheaper to maintain than terror.

The single most powerful move open to Silo 1 would not involve pipes, voice modulators or cryo-sleep. It would be offering residents a believable future: a real, verifiable date when the doors will open, and a reason to wait for it. People will put up with a great deal for something they believe in. Silo 1 chose fear instead, and with Juliette heading for its front door in the final season, it may be about to find out how poor a choice that was.

Wherever I go, there I am, I am, I am…