Bottom Line: Louis Rigaud's card-battler about fighting your own depression is one of the sharpest, most emotionally honest roguelikes of its generation — but the iOS port is a broken promise, and the Switch and Steam versions are the only ones that honor the work.
The Gameplay Loop
The core turn is deceptively simple. You have a hand. Enemies are stacked in columns above you, marching down. You play cards to kill them, blunt them, or reshape the board. Then they move.
What makes it sing is the double-jeopardy resource model. In most deckbuilders, your deck is a renewable engine — it shuffles, it cycles, it comes back. Here, cards are finite. Every strike you throw is a strike you don't have later. So the tactical question stops being "what's my optimal play this turn" and becomes "what's the cheapest way to survive this turn so I still have something left for the next one." That's a fundamentally different, more anxious calculus, and it's the single best idea in the game.
It also means the failure states are legible. You rarely lose to a dice roll. You lose because you spent too freely eleven turns ago, and you can feel the debt accumulating in real time. That's the mark of a well-tuned roguelike: the death is instructive.
The lane system deserves credit too. Because enemies come in columns, threat is spatial rather than abstract. You can see the pressure building in a specific place. Deciding to let a lane fester while you clear another is a real, tense choice with a visible price. It gives the combat a physicality that pure numbers-on-cards designs never achieve.
Where the Depth Runs Out
Here's the honest part. This is not a game with a thousand-hour tail, and its own community says so. The most common criticism from otherwise enthusiastic Steam reviewers is limited long-term depth and fairly short run length. Those complaints are correct.
Fifty-one cards is a lean pool. Once you've internalized the two or three archetypes the specialization system nudges you toward, the strategic space compresses fast. The difficulty curve is genuinely well-graded — accessible enough that newcomers to the genre won't bounce, demanding enough that veterans have to actually play well — but a good curve isn't the same as an inexhaustible one. You will hit the ceiling.
I'd argue this is a feature more than a bug, and the reason is the story. The memory fragments give this thing a destination. Endless roguelikes are all middle; Iris has an ending, and the ending is the point. Judging it against a genre built for infinite replay is a category error. It's a 15-hour experience that respects your time and then lets you go. That's not a failure of ambition. That's a different ambition.
But if you're buying this expecting Monster Train's build-crafting depth or Balatro's numeric absurdity, you'll be disappointed, and you should know that before you spend money.
Onboarding and Readability
Rigaud nails the thing most indie deckbuilders fumble: teaching without tutorializing. The board state is readable at a glance. Card effects are terse. There's no wiki-diving, no keyword salad, no forty-minute explainer video required before turn one. The onboarding friction is near zero, and it achieves that without dumbing down the tactics. That's genuinely hard, and it's the strongest evidence that a single designer with a clear vision beats a large team with a feature list.