One day, Qeynos would be one of the largest cities in Norrath. Docks would facilitate trade between the coastal city and the many towns, villages, and trading camps sprinkled along the western coast. Guilds of every size would set up shop. Tradespeople would come from far and wide to buy and sell wares. The elite status of the city guards would be of such renown that they would be sent to fortify stations as far north as Karana.

Brandan McDonald and everyone else at Sony Computer Entertainment America—now Daybreak Game Company—had big plans for Qeynos. One day it would flourish into a grand metropolis.

Today was not that day.

“It wasn't very far along at all,” McDonald said of EverQuest when he joined SCEA in 1998. The Qeynos’ docks were skeletal. The game had archetypes for male and female characters, a couple of swords and shields they could hold, a fire beetle to fight, and a blue-colored moving map meant to represent the sea.

“I guess you'd call it a vertical slice today, just showing the functionality of chatting, fighting, characters walking around,” he continued. “Truthfully, I had no clue what I was working on. I was just making dungeons and items. I didn't really see the game come alive until much later.”

McDonald enjoyed putting his head down and just making stuff. At 18, he had landed an internship at Sega of America’s BlueSky Software studio through his brother. Starting out, he’d helped novice artists learn the ins and outs of modeling, texturing, and exporting 3D models, something he’d been doing for years as a hobby.

He learned plenty in exchange. Some tricks and techniques, he gleaned from his coworkers. Others he worked out on his own. “There were just a lot of cool people who worked there and who were really friendly to me and helped me out a lot. I caught on really quickly. I'd read manuals. I wouldn't really ask questions to get people to [tell me how something worked]. I just figured things out for myself. They caught on to that, and wanted me there to help teach 3D.”

McDonald had assembled his own pipeline to import art assets into Vectorman, a 2D action platformer for Sega Genesis made from 3D models, and viewed by many as Sega’s answer to Rare’s Donkey Kong Country on Super NES. Besides building a pipeline for the game, he had learned how to do collision artwork and lighting. “They had me do all this stuff that nobody else wanted to, because I was a kid.”

In 1997, BlueSky completed its contractual obligations to Sega and, for a litany of reasons, began to fall apart. Several friends of McDonald’s landed on their feet at Verant and talked to higher-ups there about bringing him on board. McDonald’s interview was straightforward. “They talked to me: ‘Hey, do you ever play Dungeons & Dragons or MMOs?’ I said, ‘I've played Dungeons & Dragons, but I've never played MMOs. I don't even know what that is.’ They were talking about a MUD. I didn't know what that was.”

His technical acumen won over lead developer Brad McQuaid. So did one of his hobbies: Collecting swords and knives. He’d been to Germany, where he had snapped pictures of castles and brought back various types of blades as souvenirs. “I interviewed with Brad McQuaid. I talked with him for half an hour and he said, ‘Yeah, you're in. Can you start next week?’ I said, ‘Can I start tomorrow? Let's do this.’ I was only out of a job for two days before I started working on EverQuest.”

Verant’s team was plucky when McDonald joined. Their recruiting practices consisted of canvassing nearby college bulletin boards with Post-It notes advertising jobs. Most of the dozen or so developers on the team had never made a game. They bonded quickly over their excitement for EverQuest and over extracurricular activities that took some of the sting out of long nights and weekends spent at the office. Out of the blue, someone would show up with a kegger. Sometime later, catering would show up with food.

“We'd just have parties for whatever reason at all,” said McDonald. “It was fun. We were just having a good time. The culture wasn't... very corporate. That's how I'll describe it. It was very relaxed,” McDonald remembered, laughing.

McDonald put his versatile artistic skills to work designing levels. He quickly hit a snag. The level editor used at SOE was slow and archaic, lacking basic functionalities such as copy-and-paste, the absence of which added hours to the construction of a single area. He pushed back against the editor for three weeks before going to Rosie Rappaport, EverQuest’s art lead, and asking if he could switch to LightWave, a software package he’d been using for years.

There was another problem, he explained to Rappaport. The tasks he was assigned weren’t complex enough to show what he could really do. He enjoyed making items, but he wanted to imbue them with personality: weapons, tribal iconography—objects that would make dungeons and other zones pop. Rappaport and McQuaid gave him the OK.

“No one wants to do something they're not really interested in, right? Because their heart's not in it,” he reasoned. “Start with where your heart is and go from there.”

McDonald set out to leave his mark on every item and doodad he made. His first item, a sword, was a curved blade with a skull for a pommel. The skull was ornamental, but distinctive. On another sword, he applied a metallic gleam that ran down the blade.

Every day he’d crank out between six and eight inventory icons—graphical tiles that represented an item carried by players—by modeling them in 3D and paring them down, down, down, until they were simple enough to squeeze into the game yet boasting enough detail that they looked both unique and communicated their function to players on sight. By the time the game shipped in March 1999, he’d made approximately 500.

“I was doing whatever,” he said. Anything a level designer needed, I made for them.”

McDonald thrived on collaboration. Nearly every day, he’d have hallway meetings with a designer. He’d bump into someone, and McDonald would break out pencil and paper to take notes. “Each designer would have a turn to talk to me about the objects they wanted. A lot of it was the basics: lighting, such as torches and braziers; tables and chairs.”

Tables and chairs were utilitarian and necessary, but McDonald probed for more. Did the zone need a throne? And if so, how would should it be to accommodate the race of the ruler that sat it? What should adorn the seat? Oh, there’s a seamstress in the area? What type of clothing does he or she wear? Do they make their own? What’s going on in the area, anyway? Is it a slum? A warn-torn district? An affluent human city with paved streets? And if so, what type of paving material?

“I really liked asking, ‘What's the history of this zone? What's the story?’” McDonald explained. “Maybe dwarves lived there, so there's dwarven architecture, but then goblins took over, so that means adding goblin trash on top of it: mess piles and stuff. Sometimes designers will give you an object list, but you'll ask, ‘Okay, but who's this person's deity?’ I need something to associate everything with.”

Once he had an idea for a zone’s setting and story, McDonald dived head-first into research. He’d look up blacksmithing websites and extrapolate from images he found. On other days he’d read about Viking burial grounds to study photos of old swords and armor.

For the most part, the sky was the limit. “Brad wanted things more realistic,” recalled McDonald. “He didn't want anime-style, over-sized swords. He was really against that. I'd tried to put in some gigantic swords, and he was just like, ‘No.’ So I'd scale things down. I just tried to pull stuff out of history and put an EverQuest spin on them.”

Technical limitations gave McDonald parameters to work within. Most swords and shields had to be made from between 15 and 20 polygons, 25 at the most. Higher image maps were superfluous anyway, senior engineers explained, because even if McDonald managed to cram in gorgeous textures, the engine wouldn’t be able to render them to their full potential.

McDonald walked a thin line between resources and readability. A dragon statue had to look like a dragon without blasting hundreds of polygons onto the screen. “Mainly I would work on profiles, try to get a profile to look like what it was supposed to look like,” he said.

Looking at the profile of a run-of-the-mill dragon, McDonald scrutinized key features such as wings, claws, and the head. Other body parts were consolidated to save polygons. “I'd put its legs all together. Its front and back legs are consolidated, merged together. I'd position it in a way to consolidate parts, but also make it in such a way that, from far away, you'd see its profile and say, ‘Okay, yeah, that's a dragon.’ I was pretty good at looking at something and just figuring it out after a while. I'd been doing this since I was 14, so I could put things together in my head. It's not easy, and it took me a long time to learn, but it's all about profiles.”

Building an object wasn’t as direct as researching a weapon or piece of furniture and painting it. First, McDonald modeled the item in 3D. The type of item dictated the number of polys he’d use. His dragon statue contained around 200, because it needed to have more detail than a sword or shield. Then he’d map textures to the model and export it using 3D Studio Max.

To put the object into the game, he had to reference it in a definition file. “We had this gigantic text file of all these object definitions, and I had to do this for each object. Eventually they automated that process, but for a while it was super-hard to get art into the game.”

EverQuest’s object definition file spelled out information such as shaders, text that referenced colors and textures that should be applied to an object, surfaces, and other models. The programmers told McDonald to copy this, change that name… He nodded along, bewildered at first. He was no engineer, but he figured things out.

At first, he attempted to drop dozens of items into the game at once. That meant navigating a labyrinth of changes and calls to the definition file. He eased back, building out two or three objects at a time. By the time the game entered its final stretch of development, he was doing batches containing as many as 20. “We started splitting up object builds into batches, and we'd have only 400 objects per batch. It would take two to three hours to do an object build.”

EverQuest seemed to enjoy tormenting its creators with obtuse or nonexistent error messages, punishment for a lack of forethought as to the explanation of those errors. According to the level editor, vis calc was short for visible calculation error. In English, that meant rooms with too much detail would cause the editor to crash. “We'd have to go in and rip out [architecture], figure out new measurements, rebuild pillars and stuff as objects,” McDonald said.

Vis calc errors were almost guaranteed in areas that were too massive, because the engine was forced to render a large amount of polygons from the player’s vantage point. A long hallway with doors, for instance, with each door leading to a big room full of objects, was taboo. “Technically, if you have too much stuff showing, you have to have mountains occluding it, and making things a certain way so the level will run well,” explained McDonald.

If a map shimmered in-game, a pixel somewhere in the architecture was shifting back and forth. Maps had to be blurred to prevent shimmering. “You couldn't add too much detail to certain maps because they'd flicker and look weird, like glitter,” remembered McDonald.

Some tasks were as amusingly ridiculous as they were tedious. Early in development, placing an object required developers to enter the game—not the editor, but the game client itself. New objects would spawn at location 0,0,0, the center of the nascent world’s three-dimensional space. To place it where it was supposed to go—such as a throne at the far end of a throne room—McDonald or another artist would have to manually move it by steering with arrow keys, like movers lugging furniture into an empty house.

“Imagine every little object, and how big the zones were,” he said. “Dungeons had multiple floors. You're steering, like, forward arrow, forward arrow, on and on, left arrow, on and on. It was just ridiculous. We could hold down an arrow and it would take down the hallway. I wish I was drunk while doing it because it's something you could easily do while drunk.”

Eventually, programmers expedited item placement by stuffing new objects into an avatar’s inventory and sending someone into the game world to drop them at precise locations. “You'd run around and objects would pop out of your character as you created them, populating the room.”

Some bugs reared their heads only in the most specific of circumstances. On one occasion, the artists and designers wanted to light up a wall made of plaster and lined with pillars. The centerpiece would be a tapestry. On export, however, the tapestry appeared halfway underneath a pillar or in the corner of a wall, so that it was half in the level’s geometry and half outside it. McDonald or another artist performed trial-and-error placement, exporting it first a little more to the left (too far), a little more to the right (not far enough), until, at last, it appeared in the right spot.

“There were lots of workarounds, McDonald said.

A few hours to export objects was no time at all within the context of building out entire zones. Two machines in the office were devoted to the effort. They ran constantly, and each machine took approximately six weeks to perform calculations and spit out a finished, playable level. Sometimes a machine would encounter an error and halt a few hours in, or a few weeks. There would be no error message. McDonald or another artist would have to pick their way through the zone, object by object, file by file, to determine what had clogged the gears. They would remove the offending part, and the six-week process would begin anew.

“The machines were just sitting there running all the time. It was ridiculous,” McDonald admitted, “but that's how we had to make the levels.”