Jeff Kaplan: World of Warcraft, Overwatch, Blizzard, and Future of Gaming | Lex Fridman Podcast #493
Summary
Kaplan’s clearest operating thesis is that breakout games come from small, trusted teams with creative control—not anticipatory hiring or revenue targets. Titan consumed $83 million, grew beyond 140 people, hired roughly 70 environment artists in one year, and left a key technical artist able to use its editor only 20 hours in a 40-hour week; the surviving team built Overwatch by reversing that model. Its governing phrase became: “The best feature we can add for the player is shipping.”
World of Warcraft expanded the MMO market by making efficient progression follow the world’s quests. After a play test exposed that players expected uninterrupted quests, Blizzard loaded experience into quests so the “path of least resistance” moved people through the world and its stories. Kaplan links WoW’s leap from EverQuest’s hundreds of thousands of users to tens of millions partly to letting people play alone inside a populated world: directed gameplay that “felt optional, but really wasn’t.”
Blizzard polish was an organizational capability spanning embedded QA, trusted specialists, and software designed for rapid hotfixes. QA systematically covered regressions and hardware configurations, maintained raid teams, and included shooter specialists who could detect a frame of input lag; server architecture could ideally restore a disabled Overwatch hero within half an hour. The commercial benefit was felt as craftsmanship: players notice whether developers “clean up their fucking yard.”
Overwatch’s live service caught demand, but competing commitments converted momentum into organizational drag. Summer Games, Lúcioball, and Halloween events validated Kaplan’s “catch the wave, ride the wave” instinct, while a PvE Overwatch 2—discussed from 2015—and Overwatch League pulled resources away. The league’s player protections and city-based teams were defensible, but pitches implying NFL-scale success, billionaire franchise investors, a 2018 Twitch deal, and extensive integration obligations created “a house of cards waiting to fall.”
Kaplan’s departure shows how financial governance can rupture a high-performing creative organization. He says a CFO imposed an unspecified recurring-revenue target and warned that missing it would put 1,000 layoffs “on you,” amid suggestions that matching Fortnite’s roughly 1,400-person staffing would reproduce its economics. Blizzard had shifted from about 95% developers and 5% operations when he joined to roughly 50/50 across 4,500 people when he left; his warning is blunt: developers are “the golden goose.”
Kintsugiyama is a deliberate experiment in capital efficiency and founder control. After roughly 60 parties approached Kaplan in April 2021, he and Tim Ford calculated the team, schedule, and budget their game required, raised only to that level, and optimized for control rather than auctioning the company. The resulting 34-person team is building The Legend of California, an open-world online action game set on a mythical Gold Rush-era island, with a public-ish alpha targeted for March and an intentionally rough Early Access period.
Kaplan designs retention around emotional contrast, changing motivations, and the stories players generate for one another. EverQuest held him for roughly 272 played days in three years—more than 6,000 hours—and ultimately supplied both his career and family; Rust has held him for about 5,000 hours through monthly resets and the possibility of catastrophic loss. Yet he would make a new hero shooter less team-centric than Overwatch, because players experience six losses differently from a preceding eight-game winning streak and naturally seek evidence of individual contribution.
Kaplan sees AI as a useful production tool but not a substitute for authorship, making small studios his core industry bet. Today’s systems are, in his experience, confidently wrong and sometimes “right one in 10 times,” though automating a tedious resize of 2,000 images can help without displacing a planned hire. Permission remains his ethical line for art, voice, and other creative work, while his categorical forecast is that “small studios are the future of gaming” and human creators must “own the craft.”
Deep dive
1. Cheap machines taught Kaplan that imagined worlds beat fidelity
Born in 1972 and raised in Southern California, Kaplan entered games through the coin-op culture of Pac-Man and Asteroids. His Uncle Ronnie supplied quarters, while Kaplan and his brother chased an Asteroids high score only to find someone had beaten it overnight.
Pong made playing through a television feel impossible; Intellivision, supplied through his father’s relationship with Mattel, became the home obsession. The standard was whether a console game approached the arcade, until the NES and Super Mario’s hidden secrets showed him “the world opened up more and games could be more.”
On his parents’ IBM, Zork replaced graphics with commands such as “Go west. Open mailbox.” Kaplan argues its imagined world explains why “the book is always better than the movie”; when a later CD-ROM Zork supplied graphics, it shattered the private version players had constructed in their heads.
2. Online shooters made other humans the magic
Ultima II showed Kaplan a world rich enough to contain fantasy combat, merchants, bouncers, robbery, attempts on Lord British, and even a rocket ship. He missed Ultima Online during graduate school, but its poisoned-apple griefing stories revealed an early multiplayer “ant farm” where humans were the experiment.
Doom first made a modem-connected opponent feel magical: “That’s a person on the other side of that.” Quake refined the experience into a community of mods and websites, split between “low ping bastards” and everyone else; Kaplan happily entered matches through dial-up at roughly 300 ping.
Blue’s News and developers’
.planupdates connected that community to its creators. When id programmer Brian Hook announced he was leaving for Verant to build EverQuest, Kaplan reasoned that anyone leaving “the greatest institution in all of gaming” must know something—and followed the game.
3. Writing failure forced Kaplan to destroy an identity
Writing first attracted Kaplan through transportation and emotional discovery, but creation required the ego to think that after reading 1984, he could make something another person should read. His broader claim is that creators are wired to attempt it: “They can’t help but do it.”
He left a supportive NYU writing circle for California and a relationship that ended within roughly two months. Despite recognizing the move as an “epic life mistake,” he built three years around eight hours of daily writing, taking 5:00 a.m. East Coast recruiting assignments to protect the workday.
Literary submissions then required printed manuscripts, return envelopes, postage, and a cost for every submission. Kaplan received more than 170 rejection letters in one year and survived emotionally on scraps such as an editor writing “showing promise,” while isolation and alcohol deepened his depression.
Eventually he threw away manuscripts, personal journals, and roughly 20 handwritten volumes of notes. The dumpster was his boxer’s towel: keeping anything would invite another attempt in 10 years, while destroying it admitted, “I’m not going to win this fight, and I need to move on from it.”
4. Recovery began by choosing activity over status
Kaplan resists the universalized instruction to “never give up.” A collection of literary editors was telling him that writing might not be his calling, and the hardest part was closing that identity before he knew what could replace it.
His advice to young people is to stop answering adults’ questions about occupational status: “Focus on what you wanna do, not what you wanna be.” How someone spends unobserved time—and what brings fulfillment moment by moment—will reveal what they become more reliably than acting out the image of a writer, astronaut, or game maker.
Recovery included parental support, sobriety, and therapy; Kaplan stresses that the useful intervention is finding the right therapist, not checking a box. Antidepressants did not work for him, and electroconvulsive therapy became an extreme “there’s-nothing-else-to-lose” intervention that helped break the depression, not a miracle he recommends generally.
5. EverQuest turned escape into a new life
When EverQuest launched in 1999, Kaplan transferred the time previously reserved for writing into a halfling rogue in Norrath. Its loop—kill for experience, gain loot, become strong enough to kill more—made him live around the next login.
The
/playedcommand eventually showed roughly 272 days in three years, or more than 6,000 hours. Kaplan is candid that the escape was not wholly healthy: games can supply companionship and relief, then worsen withdrawal until the player needs the game to treat problems the withdrawal helped create.At level 50, then EverQuest’s cap, he joined Legacy of Steel, the leading guild on The Nameless server. Raids brought 30 or more people against dragons such as Nagafen and Vox; perfect coordination produced “high highs and low lows” strong enough to make solitary players cheer aloud or feel genuinely depressed.
6. An Uber Guild became Kaplan’s first management school
Coordinating an online guild was “like herding cats” because members wanted different things: loot, collective success, companionship, or escape. Kaplan might repeat “Everybody get to North Ro” for six hours and “like 200 times” merely to get perhaps 20% of the group there.
After founder Dread disappeared, the wood-elf warrior Ariel took over, promoted Kaplan into raid leadership, then gradually delegated more. Running raids became Kaplan’s first leadership laboratory: when to inspire, discipline, encourage, or force people who could leave at any moment.
EverQuest identities were guarded so closely that players assumed everyone was male until surprised otherwise. Kaplan met his future wife there, then gained his career through guild connections; to people claiming an MMO cannot be won, his answer is definitive: “My whole career and my family are thanks to EverQuest, so I think I won the game.”
7. Blizzard recruited a guild leader for six months without saying so
Duke Nukem and Half-Life shipped with editors, letting Kaplan build levels before he had a development job. When Ariel asked to see them, dial-up transfer would have taken about 12 hours, so Kaplan burned a CD and mailed it to an Irvine address.
Ariel revealed himself as Rob Pardo, then Warcraft III’s lead designer; several other guildmates also worked at Blizzard. Lunch with them became Kaplan’s “coming out moment”: after hiding his gaming life, he could openly discuss dragons, raid tactics, and guild drama and “felt like myself for the first time.”
Another guest, the troll Barfa, proved to be Blizzard co-founder Allen Adham. Kaplan had once saved Barfa in EverQuest’s dungeon The Hole by donating an expensive teleport potion while relying on his rogue’s stealth to escape—a player action that became part of his unannounced interview.
Pardo eventually told him to check Blizzard’s job site, where an associate quest-designer opening unusually sought a creative-writing degree. After a formal six-month process—including Chris Metzen, Kevin Jordan, Bob Fitch, and a final interview at an ARCO–Jack in the Box—Kaplan concluded, “These are my people.”
8. “One of us” entered a developer-first Blizzard
In an April 18, 2002 post, Kaplan promised that his WoW role would oppose tedious key camps, empty time sinks, luck-over-skill rewards, and rewards misaligned with effort. He hoped players would take comfort that “one of us is on the other side of the fence, looking out for the interest of the player.”
Kaplan now hears ego in that declaration: an inexperienced designer presumed that extensive play revealed how to fix the genre. Fridman’s counterpoint—worth keeping—is that the same ego and player knowledge supplied the courage to try, even though Kaplan did not yet understand implementation uncertainty, trade-offs, or how designs change at scale.
His offer was $35,000 annually, and Blizzard had fewer than 200 employees in May 2002, including Blizzard North. The Irvine office felt like a dorm: black shirts, shorts, skateboards, posters, futons for sleeping, and two main development groups—revered RTS veterans on Team One and repeatedly restarted Team Two building WoW.
9. Small teams keep every craft in the same conversation
Kaplan divides game production into five core disciplines: engineering, art, design, production, and audio. Design itself spans systems such as loot probabilities, authored quests and narrative, encounters, and generalists who connect several domains.
Audio and lighting are the two crafts players often notice only when absent. A lighting artist works alongside graphics programmers on shaders and visual targets; an audio programmer may build hooks for sound designers, translating artistic intent into functioning technology.
On a 10-person incubation team, everyone hears every consequential discussion, including server architecture and art style. Growth creates compartments and alienation until disciplines say, “The artists just don’t get it”—a claim Kaplan calls both false and demeaning to people whose life’s work is equally invested in the game.
His Overwatch assumption was that Blizzard could recruit the best prop artist in the industry; therefore leadership should treat that person accordingly. When such an expert challenges X in favor of Y, the productive starting question is not how to defend X but, “Why don’t I listen to it?”
10. Leadership improved when Kaplan stopped needing to be right
As a young WoW lead, Kaplan’s insecurity made being correct feel central to the job. In one cross-disciplinary meeting, he politely but systematically explained why others’ suggestions were wrong and substituted his preferred design.
Pardo pulled him aside afterward: “You’re a very smart designer, but you shouldn’t do what you just did to those people.” His instruction was to listen and try to make their ideas work, shifting the leader from gatekeeper to developer of other people’s thinking.
“How could we make it work?” became a core practice rather than a ritual Kaplan privately disbelieved. Some of his best designs emerged after an initial no; when they succeeded, the originator received the credit and visibility that his younger, ego-driven leadership would have prevented.
11. WoW made the world—and faction identity—the product
Metzen’s governing idea was that “the lead character of World of Warcraft is the world.” Azeroth needed to feel dangerous yet comfortable, beautiful yet unsettling, and large enough that crossing it could take half an hour; character progression served living inside that shared place.
Allen Adham’s Horde–Alliance split made every player immediately part of a team and turned affiliation into identity strong enough for real tattoos. Kaplan and Pardo initially opposed it because EverQuest allowed mixed-race groups, while Adham cited Dark Age of Camelot’s factional magic.
When Adham retired roughly nine months to a year before launch, Pardo inherited the design and reconsidered: if someone as capable as Adham had fought that hard, the team should honor the judgment. Pardo and Kaplan changed their minds and committed fully rather than reopening the decision.
The team itself was a “Bad News Bears” mixture of Blizzard veterans and newcomers. Quake-community hires initially built MMO dungeons like Quake maps, with WoW at one point using QERadiant before moving to proprietary technology; Kaplan was, by his description, “some fucking idiot” who had played enormous amounts of EverQuest.
12. WoW’s underdogs paid for ambition with extreme crunch
The competitive frame was intimidating: EverQuest II appeared to be the obvious successor, while Star Wars Galaxies paired the Star Wars property with Ultima Online veteran Raph Koster. Inside Blizzard, Team Two feared, “We’re fucked,” even as the broader company brought confidence from Warcraft, StarCraft, and Diablo.
During Warcraft III’s gold master, Kaplan spent 30 consecutive hours replaying a cinematic to reproduce a rare crash for an engineer’s logging. He watched colleagues leave, sleep, change clothes, and return while he remained seated with red eyes in the same position.
WoW’s planned early crunch slipped into an “endless death march,” remembered through catered Newport Rib and Panda Express, stained carpets, couches, and sleeping developers. Kaplan calls the environment unhealthy, even though he personally chose holidays, Christmas, and beach-to-office work rather than feeling coerced.
Fridman argues that voluntary crunch can still become a career’s most fulfilling period; Kaplan agrees while drawing a hard line around force. Executives working 40-hour weeks should not order others to stay late, but Kaplan’s normal day has rarely been below 10 hours because the work itself gives him pleasure.
13. Questing won by making exploration the efficient choice
Adham decided quests should matter, then hired Kaplan and Pat Nagle to build the system, with Eric Dodds shaping its interface. Kaplan took Westfall, Nagle took Elwynn Forest, and both translated Metzen’s stories into player movement through their zones.
Their initial plan merely matched their estimate of EverQuest’s total quests. A play test exhausted Elwynn’s dozen or 20 quests within an hour or two, and non-MMO colleagues demanded continuous objectives; the designers had an “oh shit moment” about how drastically they had underestimated content needs.
Blizzard then designed along the “path of least resistance.” Rather than let the fastest leveling strategy remain repeatedly killing easy, fast-respawning creatures, it overloaded experience into quests, making efficient progress carry players across the world, expose authored content, and enable longer storylines.
Kaplan believes this helped move the market from EverQuest’s hundreds of thousands of players to WoW’s tens of millions. Other people made the world feel alive, but forcing interaction repelled many users; WoW let them play almost as single-player players inside an MMO, with direction that “felt optional, but really wasn’t.”
14. Great design rotates players through several kinds of fun
Designers cannot rely on how players describe themselves, because people often report the gamer they wish to be. Kaplan considers extensive play essential—including bad games, where a strong system may be hidden by poor tuning, interface, or visceral execution—and studies how his wife and children respond differently.
His “quack psychologist” model distinguishes intrinsic from extrinsic motivation. Overwatch loot boxes were designed by a game designer, but their attraction combined the spreadsheet question—whether the item was useful—with the lizard-brain pleasure of sound, animation, and objects spilling out.
Kaplan’s unwritten elements of fun include progression, investment recognition, mastery, raw skill, creativity, and customization. A growing gold balance rewards accumulated investment; an appearance expresses identity; a novel Whirlwind Barbarian build rewards authorship and mechanical experimentation.
Those motivations change inside one journey. Kaplan may seek a new piece of content, become loot-motivated because a dragon requires better armor, then enter creativity mode to counter a curse—cycling through equipment, skill design, victory, and renewed exploration rather than remaining one fixed “type” of player.
15. Multiplayer’s best stories are authored after release
Kaplan places single-player, cooperative, multiplayer, and massively multiplayer experiences on a widening social spectrum. Naughty Dog-style single-player games handcraft everything for one person; Left 4 Dead coordinates four known players; multiplayer introduces strangers and conflicting goals; an MMO expands the population into thousands.
Couch co-op predates the couch in multiplayer arcade cabinets such as Gauntlet. NHL 93 and Mortal Kombat turned the controller itself into a social reward—lose the match, lose the controller—showing why sharing a room can make mechanics into rituals.
Designing multiplayer resembles preparing a movie, leaving the room, and asking actors to produce the experience without the director. Kaplan loves that uncertainty: his potion for Barfa and players’ accidental loot-goblin disasters can become stronger stories than anything a writer prescribed.
He rejects being called “the anti-story guy”; he is “the anti-shitty story guy.” Red Dead Redemption 2 combines Arthur Morgan’s authored tragedy with emergent comedy—Kaplan accidentally answered “Howdy, partner” by firing a sawed-off shotgun, triggering a town chase and a train collision—while “That’s the Way It Is” still made him call the game art.
16. A hated quest made self-criticism part of Blizzard’s craft
Metzen would draw Azeroth on whiteboards while narrating its cultures, and Kaplan translated that imaginative geography into progression. A free flight between Stormwind and Ironforge carried low-level players over Searing Gorge and Burning Steppes, planting the sight of lava and dragons as a promise of danger to come.
Green Hills of Stranglethorn scattered Kaplan’s Hemingway homage into pages dropped by any creature in the zone. The pages did not stack or receive dedicated storage, so scarce bags filled with duplicates; the intended trading conversation became aggravation, Auction House work, or abandonment.
Kaplan calls it the hubris of a junior “ant farm designer,” confusing fun for the designer with fun for the player. Publicly saying, “I think I made one of the shittiest quests in the game and here’s why” made failure discussable and challenged colleagues to improve the underlying idea instead of defending shipped work.
17. Blizzard polish was an operating system, not a final pass
Polish began as a studio-wide refusal to become comfortable with defects, not the fantasy that a game would ship bug-free. Developers who played their own game felt the cost personally, while urgency determined how quickly inevitable bugs stopped reaching customers.
Kaplan calls Blizzard QA the industry’s best because testing was systematic: plans, regression coverage, compatibility labs containing many cards and machine configurations, and a complete WoW raid team. Embedding QA beside developers dissolved the rule that testers should communicate only through a chain of command.
Specialization made that trust operational. Overwatch had QA players capable of landing repeated sniper shots from 100 meters and detecting a frame of input delay; pairing that tester directly with an engineer could reveal a defect that a merely competent shooter player would never isolate.
Engineering made games hotfixable so server-side corrections did not require every client to patch. A disabled Overwatch hero ideally returned within half an hour, because one customer might play only that hero; together with rapid balance corrections, that responsiveness let customers feel whether developers “loved it behind it.”
18. Shipping WoW was only chapter one of the live service
When Pardo left for StarCraft II, Kaplan and Ultima Online veteran Tom Chilton were put in charge in early 2005 while Kaplan was still titled Senior Game Designer. Servers were overwhelmed, database talent was scarce, and Kaplan assumed the chaos was normal because WoW was his first shipped game.
Game Masters called his home landline at 3:00 a.m. when exploits erupted. During one Stranglethorn incident, a panicked GM spawned “Guardians of Blizzard”—giant infernals that death-touched anything—to contain players while Kaplan spoke quietly beside his sleeping wife.
The WoW launcher began as an anti-cheat mechanism and nearly disappeared when that function moved in-game. Mike Morhaime insisted it return because it was the company’s best communication channel; Kaplan concedes Morhaime was right, and the launcher eventually evolved into the broader Battle.net interface.
Success did not prevent disillusionment. Groups left to form Carbine and Red 5, taking leaders, animators, artists, and programmers; Kaplan declines to speak for the departing employees, but says he thought some had been promised compensation they did not immediately see. More broadly, he says games drift from their idealized inception toward a compromised shipped reality after years of cuts.
19. BlizzCon exposed how badly forums mispriced player love
After launch, Team Two saw failing servers, departures, and forums filled with lag complaints and personal attacks. That online sample convinced developers they had somehow failed despite the game’s commercial explosion.
Pardo broadened an EverQuest-style guild gathering into BlizzCon, and the first announcement initially drew little response. When 2005 tickets appeared inside the WoW launcher, they sold out instantly; Kaplan encountered “nothing but an outpouring of love” from people desperate to discuss the game and meet others who shared it.
The contrast revised his model of feedback: passion online tends toward the harshest wording because extremity wins attention, whereas face-to-face disagreement remains surrounded by affection for the shared object. Fridman’s formulation is that communities gather because they love the thing, even when their public language hides that premise.
20. Internet criticism can select great creators out of the market
Kaplan accepts some responsibility for early online edginess, though a few surviving rants exaggerated his overall persona. He now sees social platforms as games whose scoring system rewards maximum hyperbole—especially declaring that something is “a crime against humanity”—over saying it is merely good with qualifications.
One attack can persist for years despite 100 compliments. Kaplan wonders whether this creates “social Darwinism” in which surviving criticism becomes a prerequisite for art, excluding people who might have possessed Beethoven- or van Gogh-level ability without the temperament to absorb direct, continuous abuse.
His concrete example is Diablo III director Jay Wilson, whom Kaplan regards as a great design mind driven away from games for roughly 10 to 15 years by community hostility. Fridman’s pushback becomes a call to fund creative option value socially: let creators fail, look foolish, improve, and hear explicit love before negativity removes them.
21. Titan tried to build every game at once
Blizzard began discussing a WoW successor around 2005 or 2006, assuming the MMO might last roughly five years; real Titan development accelerated in 2007. The project also required a new engine, a new team, and a new intellectual property at the same time.
Its future-Earth players were secret agents with ordinary businesses and homes by day, then performed ability-heavy first-person combat by night. Animal Crossing, Harvest Moon, and The Sims informed civilian life, while GTA-style driving connected a world spanning Bay City, a version of San Francisco, Hollywood, the rest of California, Cairo, and London.
WoW realms might then have held around 5,000 people before another server opened; Titan instead sought one server and one world for everybody. That infrastructure multiplied the creative problem of making an enormous Earth coherent while the team still debated foundational rules such as whether aliens existed.
Kaplan says he knew by deep 2009 that the current game could not ship and asked Morhaime to close it in 2010 because it would “just burn money.” Blizzard finally ended development in 2013 after $83 million, making it Kaplan’s most painful production experience.
22. Titan failed because headcount arrived before vision
Kaplan places responsibility squarely on team leadership, including himself, and says Titan failed across design, engineering, and art. Individual artwork could be among Blizzard’s best, yet the absence of a unified style made the collection look as though it came from 10 unrelated games.
His semantic distinction is load-bearing: an idea is abundant, while vision is an idea with inspiration, believers, technical and design plans, an art language, and a credible production path. Titan expressed Blizzard’s post-WoW hubris—“We can do no wrong”—without that path.
Anticipatory hiring compounded the gap. The project hired roughly 70 environment artists in one year before defining the world, constraints, or tooling, and eventually exceeded 140 people; newly relocated experts arrived expecting Blizzard’s next MMO and instead asked, “What is this game?”
Leadership then invented props and locations merely to occupy talent while core questions remained unanswered. Titan Editor, or TED, was so unreliable that technical artist Dylan Jones logged only 20 usable hours in a 40-hour week—Kaplan’s analogy is recruiting elite fighter pilots and refusing to let them fly.
23. Scope belongs to creative leadership
Kaplan’s preferred incubation begins with the smallest group capable of proving the experience using borrowed engines, placeholder art, and whatever technology is available. He hires only when the existing team is nearing overtime or the project may fail to ship without another specialist.
Titan said yes to everything; Overwatch survived by saying no. A creative leader alternates between pushing—“capture hearts and minds” when the team is too cautious—and pulling when invention outruns reality: “The best feature we can add for the player is shipping.”
Scope should therefore come from game, art, and technology directors who know which fantasies need full investment and which can be cheated. A specialist might lovingly spend six weeks on one crate, but a game needing 200 crates requires leadership to ask for hours, not solve the mismatch through meetings or indiscriminate hiring.
24. A six-week reprieve separated ideas from shippable visions
After Titan, most of its roughly 140-person team moved to Heroes of the Storm, Diablo III’s expansion, WoW, or Hearthstone. A small group received six weeks and two extreme constraints: ship within two years and retain the potential for World of Warcraft-like revenue.
Kaplan ignored the revenue abstraction but treated time seriously. The group allocated two weeks to each of three pitches and had to fall completely in love with the current one, knowing it would discard that emotional investment and start over every 10 working days.
StarCraft Frontiers imagined a lone “Indiana Jones in space” prospector exploring beneath the franchise’s giant Terran, Zerg, and Protoss wars. Arnold Tsang and Peter Lee depicted him with a cigar, a boot on a Hydralisk skull, and a Medivac behind him—art later reworked into Overwatch’s McCree.
The team designed classes, progression, expansions, and a live-content model, but recognized that 40 people could not build it in two years. Kaplan estimated at least five years with 150–200 people under ideal conditions; merely deciding that a space game needed controllable spaceflight could add roughly three years.
25. One stray sentence turned CrossWorlds into Overwatch
The second pitch, CrossWorlds, came from Metzen’s vision of a seedy neutral planet at the universe’s edge: part Mos Eisley, populated by aliens, traders, smugglers, criminals, and diplomats. Keeping their home worlds offscreen preserved future expansion territory.
During a class-versus-skill design meeting, Geoff Goodman wished the team could replace six classes carrying roughly 100 abilities with 50 classes containing only one or two defining ideas. The formal meeting continued, but the sentence stayed with Kaplan afterward.
Kaplan paired that thought with Tsang’s Titan characters, distilling the Jumper into dual pistols, blink, recall, and a time bomb. The pistols drew from Kaplan’s dual G18 loadout in Modern Warfare 2; Tsang converted the anonymous class into British Lena Oxton—Tracer—while Titan concepts also yielded Reaper, Soldier: 76, and Bastion.
This became a “play to your strengths” decision. Kaplan knew Goodman could invent focused kits and Tsang could make characters people cared about, so a roster of sharply defined individuals was a credible two-year vision in a way another MMO was not.
26. Overwatch survived because the pitch promised a crawl, not an MMO
Kaplan’s original seven-page “Monetized Shooter” deck opened with League of Legends plus Team Fortress 2 and proposed free-to-play access with purchasable heroes—an idea he now calls “fucking terrible.” Ray Gresko and Metzen nevertheless immediately saw the underlying game.
Producer Matt Hawley refused to show a creative team a deck titled around monetization, forcing Kaplan to name it in the hallway. “Overwatch” came from the winning Titan-team name that leadership had ignored after pretending to hold a democratic vote—both a high five to developers and a quiet middle finger.
Jeremy Craig transformed the crude “Jeff Deck” into a pitch for Blizzard’s directors, executives, and Activision. Its crawl–walk–run strategy made the PvP shooter analogous to Warcraft I: first prove anyone cared about the universe, then consider cooperative PvE and eventually an MMO; a Photoshopped tablet supplied the obligatory but unserious mobile slide.
27. The world and heroes made simplicity feel expansive
Overwatch’s world offered “a future worth fighting for,” replacing oppressive cargo-container mazes with an aspirational tour of Earth. Santorini could be visited because it looked wonderful; Oasis imagined an optimistic Iraq, while every environment needed to remain pleasing through hours of repeated play.
Mechanically, the game honored Team Fortress 2 and arena shooters through fast movement, expressive weapons, teamwork, and what Kaplan called a low time-to-kill (TTK), while explaining that players were survivable and could take a few hits compared with Call of Duty or Counter-Strike. It began as six-versus-six, later imposing two tanks, two supports, and two damage heroes to stabilize team composition.
Tracer was the cornerstone because blink and recall were simple to learn yet immediately suggested possibilities. McCree’s six-shooter, High Noon, and the reliably timed tumbleweed embodied Kaplan’s preference for a “fucking good gun” over heroes requiring an explanation of orb inventories and retrieval rules.
Reinhardt became Kaplan’s main: a shield instantly tells players to protect others, permits teammates to shoot through it, and pairs with a Rocket Hammer. His charge borrowed the commit from Left 4 Dead 2’s Charger—once pressed, Reinhardt is a “runaway train,” making doomed charges funny and judgment-intensive.
28. Matchmaking revealed the limits of team-first design
Matchmaking is both technically difficult and thankless because players saying they want fairness often want to be slightly better and win a close match. Lex frames the goal as keeping players near a 50% win rate; Kaplan emphasizes that the situation is zero-sum, so every satisfying victory must still create a loss for someone else.
Kaplan investigated complaints directly. A player denouncing a six-game losing streak might have enjoyed an eight-game winning streak immediately beforehand without posting praise; human psychology made the same account feel systematically persecuted.
His hindsight change would be to emphasize individual contribution more. Overwatch hid the scoreboard and used medals, but losing players weaponized those medals—“I’m the top kills”—against teammates despite tiny differences; Kaplan now sees team outcome as insufficient for fundamentally individual experiences.
29. Rust proves loss and resets can create durable obsession
Kaplan calls Quake the greatest shooter, while praising Call of Duty 4 and Modern Warfare 2 for gun feel assembled from modeling, animation, effects, sound, and multiplayer flow. Overwatch director Erin Keller and Kaplan closely studied maps such as Crash and Terminal.
Rust is a procedurally generated island that normally resets monthly, with little directed play and unrestricted PvP. Players construct bases, manufacture explosives, kill one another, and permanently seize carried or stored resources—“the most PvP thing in all of PvP.”
Online raiding attacks an occupied base; offline raiding waits for its owner to disconnect. Rust players condemn the latter as dishonorable and then do it themselves, producing the real possibility that a week’s work disappears overnight and the player questions ever returning.
Kaplan has roughly 5,000 Rust hours against more than 6,000 in EverQuest. He values resets because veterans and beginners restart “naked on the beach” together; The Legend of California borrows that renewal, likely around monthly, but is not PvP-centric and will not let another player take absolutely everything.
30. Overwatch 2 pulled attention away from a live game that had caught the wave
Kaplan, Metzen, and Michael Chu began framing cooperative PvE in 2015, before Overwatch shipped in 2016. Team members who loved its heroes but disliked competitive shooters genuinely wanted another way to inhabit the universe, giving the sequel internal creative demand.
Executive pressure converted aspirational slides into commitments, particularly a claimed 2019 release. Kaplan’s warning is that a PowerPoint date shown to corporate leadership “might as well” be etched in stone, even if it was originally no more serious than the pitch’s fake mobile mockup.
The live game meanwhile discovered demand through the 2016 Summer Games and Lúcioball, followed by Halloween. Fans loved the events, and Kaplan’s instinct was “catch the wave, ride the wave”—most games never catch one, so a successful live service should keep producing heroes, maps, and events.
A faction of developers wanted those patches reduced so PvE work could begin, while executives demanded the sequel. Kaplan calls Overwatch 2 one of his largest leadership mistakes, though he rejects the claim that he abandoned PvP: every sequel map carried the instruction that the existing PvP audience had to come first.
31. Overwatch League turned a good esports thesis into a resource sink
Kaplan still believes in the league’s original ideas: regional teams, minimum salaries, player protections, and making esports competitors first-class professionals. The players, league employees, and team organizations were often talented and deeply committed.
The commercial pitch ran beyond that thesis. A road show reportedly sold prospective billionaire investors on a league that could become more popular than the NFL, while the original economics assumed local ticket sales and merchandise for globally distributed teams from London to Shanghai.
A large Twitch media-rights deal announced in 2018 created extensive obligations: streaming integrations, camera systems, team uniforms, and skins whose technical implementation carried memory and performance costs. World events and sequel work yielded while the game team serviced “make goods” for commercial promises.
When live events proved impractical and merchandise could not produce NFL economics, investors looked back to the game itself: “Didn’t Overwatch make like $500 million just in the live game last year?” Kaplan says demands to monetize that audience transformed a good idea, pursued too quickly for money, into “a house of cards waiting to fall.”
32. A financial ultimatum ended Kaplan’s 19-year Blizzard identity
At the time of the CFO meeting, Kaplan was given a 2020 target that he expected would slip to 2021: an unspecified recurring-revenue requirement, with failure allegedly causing 1,000 layoffs “and that’s gonna be on you.”
Other discussions reduced Fortnite’s success to roughly 1,400 employees and free-to-play distribution: hire the same number and equivalent revenue should follow. For Kaplan, this was the final signal that the institution he expected to retire from had replaced creative causality with headcount arithmetic; leaving “broke me.”
He nevertheless describes Blizzard as game-development “mecca.” Programmers Mike Morhaime, Allen Adham, and Frank Pearce had made games before running the company, understood development, and protected teams from external nonsense; Metzen, Sam Didier, and other visionaries supplied worlds colleagues felt obligated to honor.
The organizational ratio tells Kaplan’s decline story: roughly 95% developers and 5% operations when he joined, versus about 50/50 in a 4,500-person company when he left. His prescription is not anti-finance but anti-surrender: good finance people support the craft; creators should stop handing “the golden goose to people who don’t deserve it.”
33. Kintsugiyama raised for control, not the biggest valuation
Kaplan’s wife requested at least a year off after 19 years with little vacation. He pulled weeds and unlocked Dark Matter Ultra in Call of Duty: Black Ops Cold War, then found himself writing a new game in Notepad before the break had done its intended work.
Alone, he learned pieces of Unreal, Adobe Illustrator, Blender, and other tools he previously delegated to world-class colleagues. The experience restored something corporate development had buried: “the pure joy of the craft of making games,” without revenue demands or entertainment-industry machinery.
Overwatch programmer Tim Ford resigned and planned to arrive at Kaplan’s home an hour after his final exit interview. When Kaplan suggested a month off, Ford replied, “I’m a programmer. All I’m gonna do is program for a month”; the two immediately committed to a game and company.
Roughly 60 potential backers contacted Kaplan after his April 2021 departure amid intense VC, strategic, and Chinese-company demand. He and Ford estimated the required developers, duration, and budget, then refused to maximize the round: “It’s not an auction”—capital would reach the plan’s number, after which they optimized for control.
34. The Legend of California recombines a fixed world with shifting adventures
Kintsugiyama’s 34-person team is building The Legend of California, an open-world online game Kaplan calls action even though markets may classify it as survival crafting. It takes place on a mythical island shaped like California during an imagined 1800s Gold Rush.
The setting seeks period authenticity—prospectors, cowboys, guns, mines—without historical accuracy. California is an island and initially uninhabited, both knowingly false; Kaplan wants creative freedom without defaulting to the spaceships, aliens, or steampunk he might have pursued at Blizzard.
Fridman describes the terrain as voxel-based, while Kaplan explains that the recognizable island and landmarks such as Yosemite are handcrafted. Dozens of points of interest move with each seed, letting the stable silhouette host different routes and discoveries.
Difficulty currently uses four tiers rather than character levels, and their geography changes by server seed. The Mojave may be a beginner area in one world and Tier 4 endgame in another; Dread Rock, inspired by Alcatraz, might appear in San Francisco or integrate into the Mojave Desert.
35. Imperfection is both the new game’s tone and its release strategy
Albert Bierstadt’s enormous California landscapes gave the team a visual target: Kaplan sees a painting and wants to explore around its corner. Lighting artist Mike Marra translates that desire through clouds, color, sunrise, and sunset, while music and audio help turn terrain into somewhere Fridman says he immediately wanted to inhabit.
Unlike Blizzard’s “hero factory,” this world should make players lonely, frightened, curious, and small before they earn strength. Sunset is a gameplay emotion: danger should send the player toward shelter, a ranch, and a cozy fireplace until morning.
Kaplan plans to place the game quietly on Steam, target a “public-ish alpha” in March, and proceed into Early Access for an unspecified duration. The build will show more rough edges than Blizzard’s controlled BlizzCon demos, but that exposure is the point: “This is how the sausage gets made.”
Kintsugi repairs broken pottery with visible golden joins, making scars more beautiful rather than disguising them. For Kaplan and Ford, Kintsugiyama acknowledges that Blizzard gave them much but did not leave them unscarred; for development, it rejects perfectionism and finds strength in a game that is necessarily never finished.
36. Kaplan bets the future on small studios using AI without surrendering authorship
Kaplan names The Legend of Zelda: Breath of the Wild the greatest game ever made because its art, technology, combat, narrative, itemization, and physical world behave as players hope. Red Dead Redemption 2 joins his shrine for its production mastery, character development, plot, and “Tarantino-level” dialogue; Zork, EverQuest, and Rust remain personally defining.
Current AI-assisted development is “mostly a hot mess.” Kaplan finds systems overconfident and sometimes correct only “one in 10 times,” even on basic Unreal UI questions, while generated imagery often resembles “an interesting fever dream” rather than authored work.
His ethical boundary is permission: models should not use artists’, writers’, or voice actors’ work without consent. Automating 2,000 incorrectly sized images in about a minute is acceptable because Kaplan would otherwise work two extra hours rather than hire someone; Arnold Tsang’s drawing, Metzen’s storytelling, and “that human spirit” remain irreplaceable.
His industry forecast is categorical: “Small studios are the future of gaming.” Large companies obtain new ideas by acquiring small ones, but innovation begins where compact teams retain loud voices; his closing advice to developers is equally categorical: “Own our art form. Stop giving it to these fucking corporate jackals. You are the golden goose. Keep your eggs.”