Why the Bosses Are the Game
It's worth saying up front: Black Myth: Wukong is, mechanically, a boss-rush wearing the costume of a six-chapter action RPG. Game Science's first big release has crowd-pleasing scenery, occasionally inventive level design, and a soundtrack that punches above its budget — but the load-bearing pieces of the experience are its boss encounters. Roughly eighty named bosses ship in the base game. About thirteen of them carry the weight of teaching you how the system actually works.
This piece walks through those thirteen, in the order in which the game's mechanical literacy curriculum unfolds — not the order you fight them in. If a fight is on this list, it's because it teaches a specific lesson the rest of the game expects you to know.
Core Combat: Resource Layer Plus Frame Layer
Wukong's combat is two games stacked on top of each other. The bottom layer is resource management — Focus charges, mana for spells, gourd uses, and the cooldowns on your five active spells. The top layer is frame data — dodges have roughly 12 frames of invulnerability, the parry-equivalent "perfect dodge" window is about 6 frames, and most heavy attacks have an interruptable charge of around 18 frames before the windup commits.
Most players spend their first twenty hours playing only the resource layer. They build Focus, fire off Immobilize, drink. Eventually a boss — typically Yellow Wind Sage or Kang-Jin Loong — refuses to cooperate with that loop, and the player is forced into the frame layer. That moment of forced learning is the source of nearly every "I finally got it" memory the game generates.
- Smash stance is your default DPS rotation. Highest forgiveness, lowest ceiling. If you don't know what a boss is doing, you should be in Smash.
- Pillar stance grants vertical mobility and upper-body invulnerability frames. Mandatory against bosses with persistent ground AOE — Yellow Brow, Erlang.
- Thrust stance has the longest reach but the slowest startup. Late game it cracks open specific stagger windows, like Hundred-Eyed Daoist Master's second phase.
For spells, my standard third-playthrough loadout never changed: Immobilize answers nearly every "what is this boss doing" question through Chapter 3, and Cloud Step / Body-Outside-of-Body in the late game halves the duration of any high-HP, low-stagger encounter.
- Prioritize the top three rows of the Stance skill tree before touching Spell tree. The payoff is twice as fast.
- Don't hoard wine recipes for the endgame. Mid-chapter bosses spike difficulty enough that gourd capacity is often the actual bottleneck.
- If you hit a wall, the answer is usually farming Hair-Like Strands and Drunk-Worm wine ingredients, not changing your build. Stat gaps are bigger than skill gaps in this game.
Chapter 1: Black Wind Mountain
Lingxuzi, the ram-headed scholar at Guanyin Temple, is the first boss with proper FromSoft pedigree. The fight's design intent is unambiguous: teach you to dodge to a rhythm.
His three-phase attack template scales from slow to fast: phase one is purely horizontal sweeps, training "see red flash, dodge"; phase two adds an aerial slam, introducing vertical-axis reads; phase three layers chained sequences that demand consecutive dodges with no input gap between them. If you stall in phase three, it's not a skill issue. It's a tempo issue. Hit a directional input immediately after each dodge and the chain becomes legible.
Then there's Guangzhi, the flame variant. He's the first fight that actively punishes the "stay close" instinct trained by every action game you've played before this one. His horizontal flame staff has a hitbox roughly half a body-width larger than the visual, forcing you, for the first time, to adopt a mid-range stance. I'd recommend Pillar stance plus Immobilize — his stagger threshold is unusually low.
Chapter 2: Yellow Wind Ridge
Chapter 2 is where the game stops holding your hand and starts asking you to read.
Tiger's Acolyte is the textbook anti-muscle-memory boss. The third strike of his three-hit chain has a 0.4-second hesitation built in — long enough to bait every player who's been dodging on rhythm into rolling early and eating the strike clean. The lesson is explicit: stop reading the rhythm. Start reading the boss.
Yellow Wind Sage is where the chapter really proves itself. He isn't the hardest fight in the chapter, but he's the first boss whose behavior changes based on how you're playing him. Play too cautiously and he closes the gap aggressively. Play too aggressively and his second-phase Rock-Splitting Slam catches you mid-commitment. This is reactive AI — quietly, without making a deal of it — and Game Science is the first studio to ship something like this in a Chinese ARPG context.
Chapter 4: The Pivot at Lesser Western Heaven
If chapters one through three teach you how to play a Wukong-style ARPG, Chapter 4 sits you down and teaches you the same thing again with the easy parts removed.
Yellow Brow is the encounter that gets discussed in player communities not because his numbers are high — they aren't, by Chapter 4 standards — but because none of his attacks have legible windups. No red flash, no charged stance, no camera pull. Each strike simply arrives. This is the fight where the game audits whether you've actually internalized boss patterns through your body, or whether you've been responding to UI feedback the whole time.
Phase two summons six clones. New players assume clone HP is reduced. It is not — clones deal the same damage as the original, but their hitstun is reduced by half, meaning the brute-force solution that worked through phase one will now feed you to whichever clone you're not facing. The optimal answer is Immobilize plus Thrust stance for ranged punishes, then a return to Smash once the original returns.
"Yellow Brow is the first time the game stops being polite. He tells you, plainly: you haven't really learned to fight yet."
— Game Science art director Yang Qi, in an interview with Famitsu, 2024Later in the chapter, Kang-Jin Loong introduces a different test: a parry window compressed to about 4 frames, where every other boss in the game sits at 6–8. The technical implication is that reactive play has hit its ceiling — from this fight onward, the game wants prediction, not reaction. This is the moment Wukong shifts from a reflex test to a guessing game with deep subtext, and it's the single most important transition in the campaign.
The Boss Roster, At a Glance
| # | Boss | Chapter | Difficulty |
|---|---|---|---|
| 01 | Lingxuzi | 1 — Guanyin Temple | ★★☆☆☆ |
| 02 | Guangzhi (Fire Monk) | 1 — Black Wind Mt. | ★★☆☆☆ |
| 03 | Black Bear King | 1 — Lair | ★★★☆☆ |
| 04 | Tiger's Acolyte | 2 — Yellow Wind Ridge | ★★★☆☆ |
| 05 | Yellow Wind Sage | 2 — Sandgate | ★★★☆☆ |
| 06 | Captain Wise-Voice | 3 — New Thunderclap | ★★★☆☆ |
| 07 | Yellow Brow | 4 — Lesser Western Heaven | ★★★★☆ |
| 08 | Kang-Jin Loong | 4 — Cloud Court | ★★★★☆ |
| 09 | Hundred-Eyed Daoist Master | 5 — Webbed Hollow | ★★★★☆ |
| 10 | Yaoguai Chief: Wandering Wight | 5 — Webbed Hollow | ★★★★☆ |
| 11 | Erlang, Sacred Divinity | 6 — Mount Mei | ★★★★★ |
| 12 | The Great Sage's Broken Shell | 6 — Endgame | ★★★★★ |
| 13 | Sun Wukong (true) | 6 — True Ending | ★★★★★ |
Endgame: The Great Sage's Broken Shell & Sun Wukong
I've left these for the end not just because they close the game, but because they recapitulate every design idea Wukong has been threading.
The Broken Shell's moveset, frame by frame, contains direct echoes of every earlier fight: Lingxuzi's horizontal cadence, Tiger's Acolyte's delay-third pattern, Yellow Brow's no-tell strikes, Erlang's gap-closing pressure. The encounter is essentially a final exam over the curriculum the game has been quietly setting since Chapter 1.
The true ending fight against Sun Wukong is the rare boss in modern action games where the optimal strategy is, against your instincts, to refuse the punish window. The original's combo strings are roughly one tempo longer than they look, and almost every "I have time for one more swing" moment is a scripted death trigger. My third-playthrough no-hit run came down to one rule: when in doubt, two-hit punish only. The fight rewards discipline, not damage.
What the Game Was Actually About
After 142 hours and three playthroughs, my read on Black Myth: Wukong's boss design is this: it doesn't teach you how to win. It teaches you how to read. Read movesets. Read your own panic responses. Read the difference between trying again and trying differently. The early bosses install reflexes. The mid-game ones uninstall them. The endgame demands you replace reflex with judgment.
This is the first time a Chinese-developed AAA action game has stepped into the design philosophy conversation FromSoftware has owned for fifteen years. Wukong isn't perfect — animation cancellation feels ungenerous, the camera struggles with vertical fights, and the game's pacing flatlines for stretches of Chapter 3. But the trajectory is unmistakable. Game Science's first major release argues, fight by fight, that there's a way to write action games that respects the player's intelligence. The argument lands.
Next month we're publishing the companion piece on optimal Spell tree respec windows between chapters. Subscribers get it Wednesday morning. We'll see you then.