PEAK looks like an accident. Two small studios meet in South Korea, build a co-op climbing game in about four weeks, release it on Steam on 16 June 2025, and sell over ten million copies. Console ports for Nintendo Switch 2, PlayStation 5 and Xbox Series X/S follow on 22 October 2026 with online crossplay. That is the short version.
The longer version is a technology choice: PEAK runs on Unity. Developers are Aggro Crab and Landfall Games, publishers are the same two studios. The engine credit is not a footnote. It explains why a jam prototype could become a live game without a rewrite, and why daily map rotation and proximity chat work at this scale.
This article answers the question players keep typing into search — what engine is PEAK made in, and why does it matter — then adds an opinion other guides skip: PEAK wins because Unity let it treat stamina as social glue, not as difficulty.
What PEAK actually is
PEAK is a first-person co-op climbing game for up to four players. You play castaway Scouts after a plane crash. You climb a randomly pre-generated mountain through different biomes to reach rescue. The loop is simple: manage stamina, grab luggage for items, use rope spools and pitons, revive friends at altars, and do not fall too far from the group or the Scoutmaster appears.
Facts that matter for the tech story:
- Release: Steam on 16 June 2025 for Windows, then Switch 2, PS5 and Xbox Series X/S on 22 October 2026.
- Scale: Over ten million copies sold, with over 100,000 copies in the first 24 hours. The level of success has been compared to Lethal Company and Phasmophobia.
- Jam origin: Developed as part of a one-month game jam in February 2025 in South Korea, described by staff as fun, fast and wild cowboy development.
- Teams: Aggro Crab is a 13-person Seattle studio founded in 2019, known for Going Under and Another Crab's Treasure. Landfall is an 11-person Stockholm studio founded in 2015, known for physics games like Clustertruck, Totally Accurate Battle Simulator and Content Warning.
- Live model: Every 24 hours at 17:00 UTC the level environments and layouts rotate. A major update, The Final Ascent, shipped on 11 August 2026, with bugfixes and maintenance after.
- Engine: Unity, the cross-platform C++ runtime with a C# scripting API, first released in 2005 and popular for indie development.
If you only remember one thing: PEAK is not a climbing sim. It is a four-player stamina negotiation with a mountain in the background.
Why Unity instead of Unreal, Godot, or custom tech
PEAK could have been made in Unreal Engine 5. That is what small prestige teams pick now for high-fidelity 3D. It could have been made in Godot, which Landfall-style physics fans love for open-source control. It could have been custom C++, which is what a climbing purist would demand for precise hands.
Aggro Crab and Landfall picked boring Unity for three practical reasons.
1. Two studios needed a meeting room, not a debate.
A four-week jam between Seattle and Stockholm cannot afford an engine argument. Unity was the shared language. Both teams already shipped Unity-style physics comedy. Landfall describes itself as a small studio that stays small and makes physics games. Aggro Crab shipped cartoon 3D action in Unity-friendly pipelines before. Unity prefabs, asset-store luggage, and familiar first-person controllers let them start placing mountains on day one instead of arguing about renderers.
2. PEAK needs good-enough physics and good-enough netcode.
PEAK does not need Nanite cliffs or ray-traced rope. It needs: a stamina bar that shrinks when you are hungry, poisoned, cold, or carrying too much; luggage that spawns items; rope and pitons that hold; unconscious and death states that are readable in first person; and proximity chat for up to four players. Unity has shipped a UDP networking layer, state sync and remote procedure calls since the 2.0 era, and its 3D physics and cross-platform support are mature. Boring is a feature when your design risk is social, not visual.
3. The plan was always Steam now, consoles later.
Unity supports desktop, mobile, console, VR and WebGL from one project. PEAK proved the loop on Windows, then ported to three console families with crossplay in about 16 months. You do not do that from a jam codebase unless the engine already handles controllers, performance scaling, and platform certification plumbing. Unreal could do it too, but not with 11 plus 13 people and a jam hangover.
What Unity gained for PEAK
Stamina that creates conversation.
While climbing you must watch a stamina bar that loses max capacity to hunger, injury, item weight, poison, temperature and other hazards. If it empties you go unconscious, then die unless a friend restores you. That sounds punishing, but in practice it forces talk: who carries the backpack, who eats first, who places the piton, who waits. Solo climbing games test grip. PEAK tests group budgeting. Unity's simple UI and trigger volumes make that bar legible at a glance, which is why strangers can cooperate without a tutorial.
Luggage, rope, and altars as readable verbs.
Levels are divided into six biomes with rotating variants — Shore always first, then Tropics or Roots, Alpine or Mesa, Caldera/Kiln or Gloom/Citadel, then Peak — with luggage cases from the crashed plane scattered around. Items are consumable and often trade stamina for safety. Backpacks expand inventory. Rope and pitons create shortcuts. Altars at level ends revive the dead. None of this needs custom tech, but all of it needs clear pickup, throw, place, and revive interactions. Unity's component model is ideal for that kind of item soup.
A daily mountain without a battle pass.
The 24-hour rotation at 17:00 UTC is the smartest cheap retention in recent co-op. Instead of adding ranks, cosmetics grinding, or seasonal power, PEAK just deals a new mountain every day. Players return to compare todays Tropics vs yesterdays Roots, not to avoid missing out. This is a newspaper model, not a casino model. It only works because the mountain is pre-generated from chunks Unity can stream without hand-authoring a new world each day. My read: this is why PEAK passed ten million without feeling like a service game.
Proximity chat as physics.
PEAK supports proximity chat. In a climbing game that means your voice fades when you wander ~200m above the group — right when the Scoutmaster hunts stragglers. Unity did not invent proximity chat, but Unity made it easy to bolt on without rewriting netcode. The result is comedy that no script could write: one friend screams for rope while the other cannot hear them.
What Unity cost
Honesty first: PEAK is not precise. First-person Unity climbing will never feel like a hand-tuned custom rope sim. Hands clip, feet slide, rope stretches oddly, and lag makes a perfect grab look like a miss. For a precision platformer that would be fatal. For PEAK it is fuel — clips become Twitch clips — but it caps how hard the game can get before frustration beats laughter.
Second cost: maintenance. The Final Ascent update and ongoing bugfixes show the jam-to-live tax. Jam code that spawns luggage fine for 100,000 players breaks in new ways at ten million, especially across Switch 2, PS5 and Xbox with crossplay. Unity helps with ports, but it does not fix design debt. Small teams pay that debt in public.
Third cost: look. PEAK will never win a graphics duel. Low-poly biomes, simple fog, and chunky Scouts are charming, but they are also a ceiling. The game compensates with readable color — poison green, cold blue, lava orange — rather than detail. That is the right trade, but it is a trade.
My take: five ideas other guides miss
These are opinions, not patch notes. I label them as such because the task is to think, not to repeat store pages.
1. Stamina is not difficulty, it is a script.
Most guides treat stamina as a meter to upgrade. I think PEAK uses stamina to force roles. The heavy carrier becomes the planner. The hungry scout becomes the beggar. The poisoned climber becomes the priority. No dialogue tree could create that fast. If you remove stamina, PEAK becomes a walking sim. If you double it, PEAK becomes lonely. The current tuning is social engineering disguised as survival.
2. The mountain is a newspaper, not a world.
Open-world games promise forever. PEAK promises today. A rotating mountain gives friends one shared topic per day — did you get Mesa or Alpine? — without requiring 200 hours. I call this the newspaper mountain. It is the opposite of live-service FOMO, and I think it is why parents, workers, and students kept playing after streamers moved on.
3. The jam was a filter, not a miracle.
People say how did they do it in four weeks as if speed equals genius. I think the four-week jam in an Airbnb in Hongdae filtered out everything that kills co-op: skill trees, crafting bloat, voice-acted lore, anti-cheat arms races. What survived was climb, share, revive. Unity gets credit here for being boring enough that no one could hide behind tech. Cowboy development only works when the horse is tame.
4. PEAK monetizes falling, not summiting.
Friendslop is often used as an insult for cheap co-op. I think it is a business model: sell the blooper, give away the summit. No one posts a clean ascent. Everyone posts the moment a friend with full luggage drags all three teammates off a piton. Unity physics jank is not a bug in this model. It is the marketing department. A custom engine that never glitches would make PEAK less shareable.
5. Unity was the diplomacy.
Two indie studios with different humor — Aggro Crab cartoon Soulslike irony, Landfall physics slapstick — need a neutral tool. Picking Unreal would have meant one side retraining. Picking Godot would have meant porting risk to consoles in 2026. Picking custom would have meant no game. Unity was the only choice that let artists argue about biomes instead of programmers arguing about builds. Boring tech can be good management.
What indie devs can steal
Do not copy PEAK's mountain. Copy its constraints.
- Budget stamina, not health. If your co-op has a shared resource that shrinks under stress, players will talk without you writing dialogue. Start with three drains — weight, weather, wounds — and tune so one player cannot carry everything.
- Ship one newspaper system. Daily rotation beats weekly events for tiny teams. One timer, one seed, one shared seed for all players. Announce the time in UTC in-game so Discord can plan around it.
- Make failure watchable. Unconscious before death, visible altars, audible proximity panic. If a fail state is not clip-worthy in 15 seconds, redesign it.
- Prototype pickups before movement. PEAK's rope, luggage, and backpacks matter more than perfect hands. If your jam has one week, spend four days on items and three on locomotion.
- Plan the port on day one. Even if you launch Windows-only, keep controller, performance scaler, and crossplay ID in the Unity project from the start. PEAK's 16-month path to Switch 2, PS5 and Xbox would have been impossible from a jam fork that ignored consoles.
Try the ideas in your browser
You cannot practice PEAK's exact stamina inside a review, but you can train the underlying skills that make PEAK fun: route reading, grip budgeting, and committing to a line. These browser games are not clones, they are trainers for the same muscles.
If you like the only-up pressure of reaching rescue, start with Blocky Parkour: Only Up Adventure, where the mission is to ascend a blocky universe in record time. It teaches the PEAK habit of looking two jumps ahead instead of staring at your feet.
If you want pickaxe-style ascent with collapsing platforms and lava, try Block Climber, where timing and tool placement matter more than speed. It is the closest browser parallel to PEAK's piton logic: place support, shift weight, climb, repeat.
If your group keeps dying to panic grips, warm up with Space Waves, a one-button neon arcade game about feathering height through tight tunnels. PEAK's stamina and Space Waves' feathering share one rule: short taps beat long holds. Learn to stay center-line in Space Waves and you will waste less stamina over-gripping in PEAK.
Play one Classic level in Space Waves, then one ascent in Block Climber, then queue PEAK. You will feel the transfer immediately.
Bottom line
PEAK did not win because Unity is powerful. It won because Unity is predictable. Predictable physics, predictable netcode, predictable ports let two tiny studios bet everything on unpredictable friendship.
Ten million copies, a one-month jam, and a daily mountain at 17:00 UTC — that is not a tech demo. It is proof that boring engines make room for interesting people. If you are building co-op in 2026, steal that order: boring stack first, social stamina second, mountain last.
My analysis above is an interpretation of public facts and design observation, not insider reporting. Engine, release, sales, jam, rotation, and team facts are cited to their pages. The five takes are mine.
Top comments (0)