June 1, 2009, Los Angeles, Calif. — Galen Center at the University of Southern California

Behind the conference

2001, Elstree, England — Peter Molyneux's house

As Lionhead finishes work on Black & White's "Creature Isle" expansion pack, Peter Molyneux invites some 15 Lionhead employees to his home — a country house, filled with hidden rooms and board games, where Lionhead staff often go to brainstorm. He arranges the get-together after work. Some invited are Black & White staff; others are new to the company. They arrive to answer two questions that sound like one, recalls programmer Tak Fung: "What kind of game is Dimitri, and what kind of game are we gonna do?" Into the evening, the group throws out ideas for an adventure game built around artificial intelligence experiments. One team member suggests a '50s America setting, with Norman Rockwell and Edward Hopper art influences. Another suggests a Twilight Zone-style mystery for a story. Another references Twin Peaks and says they have the DVDs to pass around. By the end of the day, the team has a focused plan for the game's time period and atmosphere, but a less focused plan for its gameplay. Project Dimitri begins development.

Arguably a game and arguably part of Milo & Kate's history, Dimitri was in some ways Milo before Milo existed. Publicly, Molyneux acknowledged it as an AI experiment named after his godson, Dimitri Mavrikakis. Behind the scenes, sources describe it as an idea incubation lab loosely centered around an ever-changing product. An accountant's nightmare, where the team could be creative without worrying about the consequences. "I think Dimitri was always the codename for, 'We don't know what we're making yet,'" says Gary Carr, who would later serve as creative director on Milo & Kate. "Peter would give us some very high-level design brief, we'd go off and do something for as long as a year sometimes, and then he'd come out and go, 'No, that's not what I meant,'" says a Dimitri staffer. "And then the team would change. People would go onto other things. And then it would regroup." From the project's start in 2001, Molyneux handed off an idea: a high school student arrives in a new town, and has to choose which clique he wants to befriend. Then Molyneux left it up to the team to put that into practice. His approach meant trying to inspire the team rather than micromanage, given his increasingly busy schedule helping run the company. "That was the way that Peter worked in those days," says a Dimitri team member. "It was very organic, so you'd have just the seed of an idea. He'd let people go off and kind of play around." Dimitri was to be an "interactive drama" where players could create their own character then play out something resembling a classic adventure game brought up to date, with choices that made a significant impact on how the game unravelled. The project's first iteration focused on the '50s America, Rockwell, Twilight Zone-style setting established at the initial team meeting. But Fung says that led to a lot of domestic sequences, like the player ordering food in "some kind of LA Noire [role-playing game] thing." That sense of experimentation would be an ongoing theme.Players would stumble into "micro-stories," according to Fung, depending on where they went and what they did. In one sequence, a bully would harass the player, leading to a series of "choose your own adventure" story choices — one being to put a scoop of chili in the bully's breakfast when he wasn't looking, then run away as he fell to the floor. It showed how emotions could work as gameplay, played out through a series of interactive scenes. The team built a town that would serve as the game's hub, while experimenting with crowds, vehicles, rendering and lighting techniques, a landscape engine and ways to tell a story in that setting. But sources recall that the ideas rarely fit together in a way that would resemble a working game. As Dimitri evolved, Lionhead shuffled the deck and moved staff members between projects. And at times people on staff would experiment with ideas to try to spark the game into something on their own. Lionhead artists Mark Healey and Christian Bravery, for instance, spent a couple months on a pre-rendered mockup — inspired by Back to the Future — where a doctor sat the player in a time machine and fired a laser at him, turning him into a 1,000-year-old corpse. "We were trying to sell [the idea] of traveling back and forth in time to mess around with cause and effect in a small town, to solve various puzzles," says Healey. "Something like that." That sense of experimentation would be an ongoing theme throughout Dimitri's development. "Dimitri was this ongoing thing that no one knew what the hell it was supposed to be," says Healey, "so we were always trying to come up with various ideas to try and convince the rest of the team to focus on something."

Dimitri falls

2007, Guildford, England — Lionhead's office

At the moment, no one at Lionhead has heard of Kinect. Much of the studio is deep in work on Fable 2, and Lionhead just canceled Survivors. Former The Movies studio head Gary Carr — a longtime Molyneux associate with an art school background — fresh off a series of game pitches and prototypes, has been struggling to nail down his next big project. Molyneux calls Carr in for a chat. He has an idea he wants to try, and as Molyneux typically does, he gives Carr a high level idea and asks him to riff on it. "'I've always made games about populations, lots of people,'" says Molyneux, as remembered by Carr. "'I really want to make it about a person. So boil it right down into a simulation about one single being.'" Molyneux suggests a young child for gameplay reasons, "because a child is like a sponge — it can learn and it can develop and grow," says Carr. A team begins to form, made up largely of staff from The Movies, Black & White 2 and Survivors. Most conversations on the team don't reference Dimitri.

Milo enters development, on a controller

Gary Carr jumped on board from the start as creative director, trying to build out the concept. "It was purely about taking Peter's idea of a single being who's a sponge, and evolving a single human," says Carr. The player would serve as Milo's imaginary friend — a Big Brother/Sister of sorts, helping to raise him, teach him games and build his confidence. It was about parenting, in some respects, as much as it was about friendship. As Carr and others chipped away at the concept, a small team built a prototype where the player could wash dishes, flip through a journal, help Milo care for a vegetable patch and teach him to walk on his hands — balancing with an analog stick in the pre-Kinect days. Some say he'd been inspired by watching his son Lucas grow up.Lionhead also hired people from outside the game industry to work on a story that would involve player choices, bringing in film director John Dower and theatre/film/TV/radio writer Lin Coghlan to construct a narrative that would make the project feel like a film, despite its nontraditional interface. It started going well enough that in press interviews at the time, Molyneux began talking the project up. "This discovery has lead to us starting a game, and this game will be on the front cover of nature magazines and scientific magazines," he told website Gamers Global in early 2008. "It's that significant. I think it's such an incredible thing that we're doing, I think it is ... important."

Molyneux, as the general manager of Lionhead and soon to become the creative director of Microsoft Game Studios Europe, is one of the first in the Xbox division to get word of Kinect. Sensing an opportunity to blend it with Milo, he calls Carr over again. "'I can't tell you anything about it,'" Carr remembers Molyneux saying. "'It's this new technology coming out. I've never heard of it, and it's gonna use a camera.'" Molyneux asks Carr to attempt a version of Milo that the player can talk to. Carr starts in on ideas while paper goes up on certain office windows at Lionhead to prevent all but a handful of staff from seeing the early hardware prototypes.

The road to E3

Early 2009, Guildford, England — Lionhead's office

Due to a last minute idea from Peter Molyneux, Ralph Palmer is working late on a Friday night, touching up animations to meet a milestone for the title's E3 demo, so he can send it off to co-workers in California. At the moment, the facial recognition software registers a handful of Lionhead staff, such as Molyneux and Gary Carr, so when they walk in front of Kinect's cameras, Milo knows who they are. But it's in the early stages. The database has audio files with names for almost everyone in Lionhead's office, but they have yet to be synced up to the software so Milo can identify them. Palmer finishes up his changes for the day, then walks around the corner of his desk to Lionhead's screening room to test them before going home. He turns the system on, looks up at the screen, and hears Milo: "Hi Ralph. How's it going?" At this point, there's nothing in the software that should identify Palmer from the 60-80 other names in the system, so the code chooses a name at random and wins the lottery. It rattles Palmer, giving him "a really weird buzz" that sticks with him for days. The project doesn't work yet, but in cases like this it feels like it does.

Shortly after Palmer's run-in with Milo, Molyneux went on stage at E3 to show the world what the team had been doing. Cue the promises from the E3 video. A real live boy! A friend in your TV! For many at Lionhead, the E3 presentation put them in the spotlight like never before. Unlike with Fable — a successful franchise by most criteria — this time the mainstream media took an interest. It didn't matter that Milo didn't work perfectly. It didn't matter that many confused Kate the dog with Claire in the video. People were interested, which led to plenty of questions. Some of the biggest of those surrounded the content. Everyone knew Milo & Kate featured the player's relationship with Milo, but beyond that the waters got murky. "People tended to think Milo was a Tamagotchi toy," says Senior Technical Director Michal Doniec. "It wasn't true at all." Lionhead split Milo & Kate's gameplay into three types: on-rails story sequences with Milo, minigames and exploration. All based around the idea of the player being Milo's friend and building his confidence. The on-rails sequences resembled the E3 demonstration, with Milo leading the player around, asking questions and giving them tasks to do and puzzles to solve. In many ways, these played out like interactive cut-scenes where Milo would say his lines, then the script would give the player a line or two, and then it was back to Milo. In one of them, the player would put a basketball in Milo's room, then the two would play basketball together, and at one point Milo would kick the basketball down a hill, which led to a hole in the fence, which led to a forest, which led to Milo meeting Kate in that forest. One thing leading to another, with the player along for the ride. Breaking those sequences up, players could also search for things around the house — to find a key, for instance, which could come in handy later — and participate in minigames and activities like helping Milo with his geography homework, playing basketball, shooting paintball guns, etc. Players could buy objects like a basketball, as well, as they earned points. But the most dramatic moments came from Milo & Kate's story scenes. There were grand ideas that were too extreme to be considered for long, such as making Milo go blind.As the story began, Milo found himself moving from the city to the country, feeling lonely and not getting along with his parents. Playing up the lonely angle, players would never see any other people — Milo's parents' voices spread through the house in the background, their faces never seen. This led to Milo befriending Kate, a local dog, after meeting her in a nearby forest, and keeping her hidden from his parents who disapproved of him having a pet. For the story's climax — at least, for the version the team settled on by the end of development — Milo got upset at Kate because he thought she had been killing sheep, then he realized one night that wasn't true and headed out to find her in a storm. While in the forest, Milo got attacked by a beast — the one who had been killing the sheep — and then at the last second was saved by Kate, the spoiler being that she would die in the process. And then there were the grand ideas that were too extreme to be considered for long, such as making Milo go blind, starting the story as a grandfather then going back and time and playing through his childhood, playing as a female character named Millie and matching the in-engine weather to the real world weather where the player would be located. Some of these made staff nervous because of the amount of work they would involve. Others made staff nervous for a different reason.

Parental concerns

2009, Guildford, England — Lionhead's office

Lionhead programmer Mark Adami wants to make Milo respond like a real boy. He's been working on speech recognition, player recognition and drawing recognition with Kinect, and having the best success with the latter of the three. One of Milo & Kate's planned features involves players writing or drawing on a piece of paper, then holding the paper up to Kinect's camera for Milo to see. And Adami needs art samples to teach Milo what he should be able to recognize. Adami knows that Milo won't be able to see and respond to every drawing in the real world, so he narrows the possibilities, and sets up a scene where Milo asks the player to draw a boat. Then he asks everyone at Lionhead to draw a boat, each with their individual talents and perceptions of what a boat should look like. A pile of drawings shows up. Adami enters them into the system, one by one training Milo. Soon after, whenever someone at Lionhead draws a boat and holds it up to Milo, he guesses it correctly nine times out of 10.

Perhaps the most debated piece of the Milo & Kate story, both internally and publicly, is whether it worked as claimed. Following the E3 video, fans expressed concerns, but didn't have enough information to be able to say anything with certainty. At Lionhead, many were split down the middle on how well it worked. Some were amazed at how real Milo seemed; others worried that he was too much of an illusion. They describe Milo like a call center operator working off a script — ask the right questions, and you'd receive perfect answers; ask the wrong questions, and it could take a bit of an awkward interaction to get back on track. Sources praise Milo's ahead-of-its-time facial animation and emotions, some of the minigames like basketball and the guided story sequences. "Where I think the magic was eventually going to be found was in the story interaction elements," says one team member. "We saw real promise in being able to craft an interactive sequence around Milo, where Milo would move through the world, then stop at a puzzle which the player had to help Milo solve."

Seaman One of Lionhead's ideas to solve issues like this was that Milo could theoretically get smarter post-release — if Milo & Kate had made it to stores, it may have been able to store player voices and drawings on servers to make Milo more accurate at understanding players over time. Like Mark Adami testing boat recognition, but on a mass scale. "There was loose talk of being able to put the training software in the cloud, so Milo could genuinely learn new images," says a source. "But we knew people would just show him drawings of spunking cocks, and that would kill the magic somewhat." Ultimately, team members seem resigned that, had Lionhead finished Milo, errors would have been inevitable. They covered the design in bandages to disguise awkward moments — like Milo suggesting topics to guide the conversation and turning away from the camera when confused — so he would seem comfortable when he didn't have a proper response. But some say though the design didn't invite them to, certain players would inevitably try to see through the illusion, and that was an audience the team was unlikely to satisfy no matter how much progress it made.

Everyone asked for this story agrees that Lionhead developed Milo & Kate with an end goal in mind, that the project made it approximately 60 percent through development, that approximately 80 team members worked on it at its peak and that it could have been finished with an extra six months of production, though it would have felt rough if that happened. The public shorthand for that, however, became the question of whether Milo & Kate was officially a game or a tech demo. A question of messaging on the surface, which pointed at the bigger debate of whether Microsoft and Lionhead planned to sell it. This all started at E3 2009, when Molyneux told media that yes, it was "for sure" a game. "This is a game," he told website Edge. "It's a game that's going to be made. My ambition is to complete it when [Kinect] ships, a complete game with a story and a score and gameplay mechanics. It's not just a tech demo." But as time went on, Microsoft executives contradicted those statements. Most prominently, on June 28, 2010, Xbox Director of Product Management Aaron Greenberg told Australian TV show Good Game, "[Milo is] a tech demo and technology that continues to exist, but right now it's not a game that we're planning to bring to market." The next day, Greenberg amended his comment in a statement on Twitter: "Project Milo absolutely continues in development at Lionhead Studios; it is just not a product we plan to bring to market this holiday." The day after that, Molyneux told website GamesIndustry.biz, "Poor Aaron Greenberg — he's on the PR team; he hasn't seen it since last year, so he came up with this stock answer that Milo is alive and well and living in Guildford but it's still a tech demo. ... This is a full product that we're working on. It's not going to be released this year for sure; it's not part of the launch line-up." And two weeks after that, Molyneux appeared on stage with the project at nonprofit conference TED, marking Milo's second public appearance — this time shown live. It featured the stone skimming minigame and more on-screen icons than the previous demonstration, closer resembling a traditional game interface, and proving it could work live. Then in August, Molyneux backtracked and told USA Today's Game Hunters website, "I don't think of it as a released product at the moment. I still think this is a very, very big tech demo. I don't think of it as something that would be a boxed product on the shelf." One team member says Molyneux wanted to make it into a game but ultimately that wasn't his decision. Others give mixed messages on whether they considered it a game. "It was basically confused, and never quite solidified.""I can't say that it was 100 percent a game," says Senior Concept Artist Emrah Elmasli. "People thinking that it was a game was mainly the biggest misconception [people had about Milo & Kate] in my opinion." "I think it's fascinating that some people say, 'It wasn't a game; it was just a demo.' But it really was a game," says another team member. "I don't think we knew, to be honest," says Carr, noting that a full production game for the team would usually have more than 100 people on board. "I think we were playing with something, and we built quite a lot of it, and you could sit down for a couple hours and enjoy noodling around his house, and going fishing, and skimming stones, and doing all sorts of cool games. But I don't think we ever really sat down and said, 'This is going to be a product.'" One source refers to this as Carr "toeing the company line," while another views it as a simple difference of opinion around the office. "I'd describe Milo & Kate as a development experiment that somehow found itself in full production," says another team member. "In all honesty it really was trying to be something new and different, and there were certainly some cool and compelling aspects to it. However as a project it always lacked strong direction and clear vision ... it was basically confused, and never quite solidified into something everyone was comfortable with." None of this is to say staff at Microsoft and Lionhead didn't look into their options. Evidence that the companies were considering commercial possibilities comes from three different places — multiple sources confirm that early in development Lionhead and Microsoft held talks to include Milo & Kate as a pack-in with the Kinect hardware, that the companies discussed potentially breaking the project's post-release content into pieces and releasing it episodically and that Lionhead started work on a social game that would tie-in with the Xbox 360 project. Designer Wayne Imlach contracted briefly with Lionhead in 2010 to work on the social title, a browser and phone "lite" game known as "Milo's Projects" and "Milo & Kate for Web" that he says never got past an early prototype. The design focused on a school project where Milo would collect plants and animals in his garden, then look after them, and Imlach says it was designed to be "a companion product to the console game, so the mechanics were less geared toward the aggressive monetization and compulsive social obligation you tended to find in Facebook games at the time."

Cancelation day

Lionhead today

Illustrations: Scott Hassell

Design/layout: Warren Schultheis

Special thanks to: Ryan Gantz, Josh Laincz

Image/video credits: Lionhead Studios, Microsoft Game Studios, GameScriptWriting.com, Head In My Bag, Mega Meshes GDC 2011 presentation (PDF), John Dower on Vimeo, Sega/Vivarium