In Cell Mimic the player moves at 170 units per second. When the inspector gives chase, it moves at 178. That five percent is the most important number in the game, and everything else is built around it.
The first version of the inspector was slower than the player. It seemed obviously correct — the player should be able to escape, or the game is unfair.
It made the game boring within about two minutes.
If you are faster than the thing chasing you, hiding is a choice, and it is usually the worse one. Why crouch behind a beaker and hope, when you can simply walk away and keep walking? Every interesting mechanic — the costumes, the suspicion meter, the vision cones — became optional decoration on top of a game about holding one direction. Players do not use a mechanic because it is elegant; they use it because the alternatives are worse.
So the alternative had to be made worse. The fix was not to punish running. It was to make running impossible to win with.
The margin matters as much as the direction of it.
A much faster enemy is not tense, it is instantly fatal. Being spotted would mean dying half a second later, and the player would learn only that being spotted is bad — not what to do about it. There would be no interval in which to think.
At five percent faster, the chase has a shape. The inspector gains on you slowly and inexorably. You have several seconds while the gap closes, and those seconds are the actual game: enough time to look around, spot a wall, notice a cluster of identical props, and commit to a plan. You lose the race, but you get to choose how you lose it.
The other half of the speed design: the inspector patrols at 96, barely more than half its chase speed.
That gap exists so that the AI's internal state is legible from across the room. You never need a UI element to tell you whether you have been noticed; you can see it in how the thing moves. Slow and wandering means it does not know. Fast and straight means it does. When it becomes suspicious but has not committed, it moves at patrol speed times 1.3 — visibly quicker, not yet a sprint.
Three speeds, three states, no icons required. In a stealth game the player's most important input is the enemy's state of mind, and the cheapest way to communicate it is movement.
The inspector sees 230 units ahead in a cone about 64 degrees wide — on a stage 960 by 600, that reaches a quarter of the way across the room but stays narrow.
This shape was chosen on purpose. A wide cone makes position irrelevant: if it sees almost everything in front of it, all you can do is be behind it, and the game collapses into "stay behind the enemy". A long narrow cone makes angle matter more than distance — you can be very close and perfectly safe, or far away and completely exposed.
That inversion is where the tension lives. Standing three metres from the inspector, slightly to its side, feels dangerous and is not. Learning to trust that is the moment the game clicks.
Being in the cone does not mean being caught. What happens depends on what you are doing, and the rates are deliberately uneven:
| Your state, inside the cone | Suspicion | Effect |
|---|---|---|
| Not disguised, moving | Instant maximum | Immediate chase. No meter, no warning. |
| Not disguised, standing still | +150 per second | Fills fastest of any gradual state. |
| Disguised, moving | up to +95 per second | Scaled by your actual speed — creeping is quantitatively quieter. |
| Disguised, standing still | Zero | Completely safe, indefinitely, at any range. |
| Outside the cone: −60 per second. Suspicion only ever falls when it cannot see you. | ||
Three deliberate decisions are hiding in that table.
Standing still in the open is worse than moving in costume. This is the opposite of most stealth games, where freezing is the panic button. Here the crime is being visibly not a piece of lab equipment; once you are that, everything else is negotiable. The mechanic pushes you toward the costume rather than toward paralysis.
Suspicion scales continuously with your speed. There is no sneak toggle. Half speed produces half the suspicion, and the player can feel that relationship without ever being told it. Deliberate movement is rewarded automatically, which is more satisfying than a button that says "be quiet now".
You cannot cool down under observation. Suspicion falls only outside the cone, so waiting it out while being watched is not an option. There is exactly one thing to do when the meter is climbing: break line of sight. One problem, one answer, discoverable without a tutorial.
And the fourth row is the game's promise: disguised and still is always perfectly safe. However bad things get, there is a state you can reach that cannot lose. A stealth game needs one of those, or the tension curdles into helplessness.
When the meter fills, the inspector does not pounce. It commits to checking a specific spot, and there is a 0.6-second wind-up before the check lands.
That window is the difference between a game that feels demanding and one that feels arbitrary. It is long enough to be seen and reacted to, short enough that it never feels generous. Survive the check outside the grab radius and the inspector returns to patrol with a "Narrow escape!" — and its suspicion drops to thirty percent rather than zero.
That last detail matters more than it sounds. A near miss does not fully clear the slate, so a second scare arrives from a higher starting point. Repeated luck compounds into real danger, which is exactly what you want from a game about pushing your luck.
Once chasing, the inspector needs 2.5 seconds of broken line of sight before it gives up.
Shorter, and a chase ends by ducking behind the nearest object — the pressure evaporates. Much longer, and the game turns into a hopeless jog while a faster thing follows you forever. Two and a half seconds is roughly the time to get around a wall and into a plausible hiding spot, so escaping is possible but requires an actual plan rather than a flinch.
The behaviour the numbers produce, in order of how quickly people find it:
That progression is the whole design working. Nothing in it is taught with text. Each step is a strategy that fails in a specific, legible way, pointing at the next one.
The lesson I would carry to another project is about how to remove options.
The instinct when a strategy is too strong is to punish it — more damage, harsher penalties, a timer. That makes players feel policed. Making the strategy simply insufficient, with no penalty attached, feels completely different: nothing bad happens when you run, you just do not get away. The game has not told you off. It has told you something true about the world, and left you to draw the conclusion.
All of which came out of changing one constant from 150 to 178.
Written from the game's own source. Related reading: How Cells Actually Hide from the Immune System, and the other devlog, Godot to the Web. Questions and corrections welcome via the contact page.