Celeste's Assist Mode is a tone of voice

Two beliefs about difficulty can both sound obvious and still fight. A designer may hold a player to one vision when the game depends on it. A player may need a different vision to be in the room at all. Celeste, Maddy Makes Games, 2018, does not resolve this with a secret menu and a wink. It says, in the place where you choose, that the options exist and that using them is still playing Celeste.

Other games have drawn the line differently. Some hide the help in a cheat code, which teaches shame. Some rework the whole mode and ship it late, which teaches that the help was an apology. The special thing is the tone. The mountain is still the mountain. The game does not insult you for changing how steep it is, and it does not pretend the intended path was optional. Tone is the design. A slider without that sentence is just a slider.

Two convictions about games can each feel obviously true while pointing in opposite directions. The first says a designer is allowed to hold the player to a single vision when the game depends on it. Dark Souls does not necessarily owe anyone a conventional easy mode. Dead Rising can run its entire length on a strict clock. If weapons jam at random, that instability can be exactly the texture the designer wanted.

Decisions like those attract complaints, and they usually still have a reason underneath them. When a punishing boss, a stressful timer, a brittle gun, or a heavy penalty is load-bearing, a switch that quietly turns it off can undo the work. The rhythm of the whole game may have been built on top of that friction.

The second conviction says players should get to play in whatever way suits them. A player may have a disability. They may be new to the genre. They may be there for the story rather than the test of execution. They may love most of a game and hate one slice of it: the timers, the boss fights, the stealth, the combat. For that person, the rigid version turns something good into something they cannot finish or cannot enjoy.

Those two convictions do not reconcile on their own. Extra ways to play make games easier to get into, and they also muddy the question of what the designed experience even is. A player who changes the most important rule on the first screen may never meet the game the team spent years shaping. So the real question is not whether hard games deserve gentler settings. It is how those settings get explained.

The line that matters is not whether a player used an easier mode, an assist option, or a rule toggle. It is whether the player knew what they were changing. A game can hold one intended way to play and still open other doors for people who need or want them. The argument cools once the game says plainly which version was tuned as the main design and which settings exist to help different players reach it.

Being explicit protects both camps. Designers keep a clear statement of intent: this is the game as balanced, paced, and authored. Players get permission to step outside that intent the moment it stops serving them. Nobody is fooled into thinking every configuration is equivalent, and nobody is scolded for picking the one that keeps them playing.

That is why difficulty settings are a design problem rather than a menu problem. Wording, placement, timing, and framing all instruct the player in how to read the option. A row of modes on the opening screen says one thing. A tucked-away configuration screen with a careful paragraph attached says something else. A mode added after launch, shipped as a download, or given a name of its own each sets up a different relationship between player, designer, and game.

No toggle is truly neutral. It teaches the player how much the rule beside it weighs. Present a timer, a permadeath system, an enemy threat, or a brutal failure state as a minor preference and the player will treat it like one. Explain that the rule is part of what was authored and the player can make a real decision about keeping it, softening it, or throwing it out.

Old cheat codes made the boundary trivially clear. Type an invincibility code into Duke Nukem 3D, fly around with an endless jetpack, and nobody was in any doubt about what was going on. That was cheating. It was not the intended way through, and it still let some players reach games that were otherwise shut to them.

Player-made mods behave much the same way. Strip the turn timers out of XCOM 2, or make the enemies in SOMA harmless, and the result obviously sits outside the official design. Installing it takes a deliberate act. The player knows the intended experience has been altered, and the developer is never asked to present the altered version as an equal.

A visible boundary makes freedom feel safer, not looser. Players can experiment, mess around, dodge a mechanic that hurts, or see corners of a game they would otherwise never reach. Because the changed version sits plainly outside the main design, it never confuses the meaning of the original. Playing the game as built and bending it for your own reasons stay two different acts.

The catch is that cheat codes have all but vanished, and mods largely belong to PC players. Both also live outside the game itself. Move a similar option inside the game and the burden shifts. It is no longer a hidden string, a rumor, or somebody else's download. It is part of the official interface, so the game itself has to say what it means.

More games now ship options that can reshape the experience outright. These are not comfort settings on the order of subtitles, remappable controls, or camera sensitivity. They strip out mechanics, soften punishment, rewrite enemy behavior, or skip a challenge entirely. Handled carelessly, they can argue against the whole point of the game.

The strongest examples are specific about who each option is for and why it exists. They never claim every configuration is the same game, and they never treat the player who needs help as a lesser one. They say instead that the standard game was built around a particular kind of pressure, and that other settings are there for players whose experience would otherwise be spoiled or blocked outright.

It is a small move with a lot riding on it. A game can say we recommend this without saying you are wrong to need something else. It can say this changes the intended experience without saying you may not change it. Tone carries weight here, because difficulty settings are never only numbers. They are how a game talks to the player about competence, trust, and authorship.

Darkest Dungeon runs on punishment, stress, risk, and the occasional unfairness, and it treated its most contested rule as a deliberate configuration rather than a difficulty slider. The corpse system, added during Early Access, split the audience. Killed enemies left bodies standing for a few turns unless destroyed, which could shield the ranks behind them. Red Hook Studios thought the system was right for the game. Plenty of players thought otherwise.

Corpses eventually became optional, alongside other changes such as guaranteeing that a retreat from combat succeeds. How those options arrived is the interesting part. They were not offered on the opening screen as an ordinary difficulty pick. They sat inside a named configuration mode, and the game stated that the settings currently in force were the intended ones. Switching that mode off is what unlocked the toggles underneath.

The framing carries the message. Corpses are not an oversight or a balance slip nobody got around to fixing; they are part of what was designed. Players who cannot stand them still get a way out, and they take it knowing exactly what they are stepping away from.

Placement matters here because the corpse rule is no mere comfort feature. It reshapes targeting, pacing, positioning, and the price of every kill. Turning it off is not sanding down one rough edge, it is changing the feel of combat. Darkest Dungeon can permit that and still say, without hedging, which version the designers stand behind.

Compare that to asking a player to pick between major rule sets before they have played enough to know what is at stake. Fire Emblem Awakening puts the permadeath question to players before the game has had a chance to show why that rule has mattered to the series. The choice may well be worth offering. It also asks for a design judgment very early.

Heat Signature hit the same wall with permadeath and answered it by reframing the question. The designer did not want to drop permanent loss, because it was what pushed the game toward stranger stories and more varied play. But the useful question was never whether the game would be better without permadeath. It was whether the game could do something for players whose experience permadeath was wrecking.

The option that shipped lived in the settings rather than standing up as a headline mode. Its label said plainly that it was not recommended unless permadeath was genuinely ruining the game, because removing it meant less variety and duller stories. The language is blunt without being unkind. It states the design cost and still leaves the call with the player.

That reads better than either guilt or vagueness. The game does not need to tell players off for disliking permadeath, and it does not need to pretend the change is free. It can name what the rule contributes and when dropping it might be reasonable. Context, not judgment.

Underneath all of it is one principle: a player should never feel asked to design the game before playing it. A menu of equally weighted rule toggles leaves them guessing which build is the real one. A carefully framed assist says something more useful. This is the authored game, and here is how to change it if the authored game is not working for you.

More on this: Roguelike vs roguelite: permadeath names two rules

SOMA ran into the same problem from the horror side, and solved it with a new mode, a careful name, and a long delay. The original leans on oppression and dread. Its drowned world is meant to feel hostile while the player creeps past unsettling machines and monsters. Take away the threat and that tone is easy to break.

A popular mod proved that some players wanted precisely that. They wanted SOMA's story and its world, but being hunted was more than they could take. Frictional Games turned the idea into an official Safe Mode, in which enemies still notice the player and no longer hurt them.

The name is doing real work. Safe Mode carries a completely different attitude from wuss mode or cheat mode. It describes the function without passing sentence on the person using it. It also matches why players wanted it, which was rarely to break the game for laughs and usually to reach a version of the experience they could actually sit through.

Timing helped as well. Safe Mode landed years after release, once the standard version had already defined what SOMA was. That gap kept the launch message clean. Players and critics met the intended experience first, so the alternate mode arrived as a clearly labeled addition rather than a rival definition of the game.

BUTCHER reached for a related trick, shipping its easier mode as free downloadable content. A little distance from the main game helps signal that the option is available without being the default expression of the design. The distance need not feel hostile. It just makes the intent readable.

Celeste is a precise, charming, demanding platform game whose difficulty is the point rather than a side effect. The controls answer instantly, the levels are built tight, and the challenge is deliberate both mechanically and narratively. The climb is meant to be hard. The player is meant to fail, adjust, improve, and keep going.

It also reaches well past the main route. Strawberries pile on optional challenge. B-sides and C-sides demand sharper execution again. Celeste is not shy about being hard, it is organized around being hard.

Assist Mode sits inside that same game. It can slow the game down, hand out infinite air dashes, make Madeline invincible, or let the player skip a chapter outright. Any of those can wreck the finely tuned challenge, and designer Matt Thorson has said so. He also explained that the team decided to empower players and give them a good experience, even at the cost of some cherished design intent.

That makes Assist Mode much harder to design than a plain easy setting. It has to serve players who need wildly different kinds of help without pretending the mountain has no shape. A player who needs slightly more reaction time should not have to delete danger to get it. A player stuck on one chapter should not have to flatten the whole game. The help has to be strong and legible at once.

That is the tension Assist Mode lives inside. It is not a small adjustment; it can rewrite the entire platforming test. Celeste carries that power with unusual care, and the care is what makes the mode worth studying.

Cuphead makes Celeste's approach easier to see, because its own assist sends two signals at once. Cuphead is likewise brutal, built on boss fights that borrow from old, unforgiving run-and-gun design. It offers Simple Mode to players who want help, and the way that mode is delivered creates problems.

Simple Mode is offered before every fight, sitting right beside the normal option with no clear account of what it costs. It cuts whole phases out of boss battles, so players who take it miss some of the game's most elaborate animation and encounter design. Worse, a boss beaten on Simple Mode does not count as beaten, which means the final boss and the ending stay locked unless every required fight falls on normal.

The result is a muddled message. The mode looks like a sanctioned way through, then the game withholds completion from anyone who uses it. Help is offered and then quietly disqualified. To the player who genuinely needed it, the game seems to wave them in and lock the last door behind them.

That is a communication failure more than a balance one. Players need to know what a mode changes, what it leaves alone, and whether the game counts that route as finished. Leave the answer fuzzy and the assist produces frustration where it meant to produce access.

Celeste parts company with that approach on four counts, starting with reach: Assist Mode covers the whole game. Nothing is fenced off. The assisted route never becomes a second-class run that cannot be completed.

Second, the options come apart. The player does not press easy and receive a bundle of invisible edits. They can slow the game to a chosen percentage, add air dashes, switch on invincibility, or skip a chapter. Different players are blocked by different things. One needs more time to react, one needs the damage gone, one is stuck on a single screen. Fine-grained settings let a player remove the actual barrier instead of leveling the entire game.

Splitting the options up also protects dignity. A player takes only the help they need instead of accepting a broad label that may not describe them at all. The mode reads as a set of tools rather than a verdict on skill, which matters because difficulty support is personal. The same switch is a minor convenience for one player and the difference between quitting and finishing for another.

Third, the name is respectful. A more judgmental label was considered along the way, and what shipped, Assist Mode, is neutral and practical. It says the mode helps. It does not brand the player a cheat, a coward, or a lesser participant, and that matters because nobody can see from outside why a person reached for the setting. It might be disability, age, frustration, curiosity, or simple interest in a different part of the game.

Super Mario Odyssey borrowed the same name for its own helper mode, and it works there for identical reasons. The label frames the feature as support rather than shame. A small child, a newcomer, or somebody who finds 3D movement awkward can still take part. The goal is not to defend the pride of the hard version. It is to let more people play while the standard design stays readable.

Fourth, Celeste explains the mode before the player commits to it. The message says the mode changes the rules to make the game easier. It says Celeste was made to be challenging but accessible, that difficulty is essential to the experience, and that a first playthrough is best played without it. Then it grants that every player is different, and that if the difficulty is keeping Celeste out of reach, Assist Mode is there so the player can still enjoy the game.

Very little is left ambiguous after that. A player chasing the intended challenge understands the mode is not addressed to them. A player who needs it understands that using it is legitimate and planned for. The design is defended and the player is still trusted.

Some games are built on the premise that players may change anything, and that is a legitimate vision too. A sandbox, a simulation, or a toy-like game is often at its best when the player is remaking the rules. Other games are stronger under a tighter vision, where tension, pace, and meaning all rest on specific limits.

A tighter vision does not require shutting people out. A game can be hard, authored, and carefully tuned while still routing around the barriers that would stop somebody from playing at all. What it should avoid is giving the options that cut against the vision exactly the same weight as the ones that serve it.

Wording matters because names carry judgment. Position matters because a menu implies importance. Timing matters because an option shown before the player understands the game asks for a design decision made blind. A well-framed assist says: this is the intended experience, and here is help if that experience is out of reach or no longer enjoyable for you.

Celeste finds that balance. It hands out powerful tools without emptying the mountain of meaning. It states what the standard climb is meant to be, then lets players decide whether they need help getting up it. You can only really settle those calls in your own game by playing the build and watching where people stop, and once you have that, good information does more for players than protection does. Designers can keep them from hollowing out the game by accident, then trust them with the choice.

Most of Celeste's options are single values, which makes them cheap to offer. Game speed is Engine.time_scale. Infinite stamina and extra air dashes are values on the player that a settings autoload can override. Invincibility is a flag the damage code checks before it kills the player.

The cost is in the wording and the placement. Put the options in their own menu, reachable at any time, behind one screen that says plainly what the game is meant to feel like and that changing it is allowed. Save them per save file, so the choice belongs to the player who made it.

Then test one rule: turn every option on and play the first chapter. If a room breaks, because it relied on stamina running out or on a timing that a slower game makes impossible, fix the room. A player who needs the help will hit the same wall.

More on this: AI soulslike maker: commitment, recovery and the walk back

In the Flockbay app, build the intended hard version first. Then add one assist, a slower fall or an extra chance, and write the sentence the player reads when they turn it on. Play both. If the sentence apologises, rewrite the sentence. If the assist removes the idea rather than the punishment, rewrite the assist.

An AI platformer maker and a precision platformer template are the mountain. The sentence is yours.