Core Design: The Studio That Made and Lost Lara

Six people made Tomb Raider. That’s the number, and every time I write it down it looks like a typo. Toby Gard, Paul Douglas, Heather Gibson, Neal Boyd, Gavin Rummery, Jerome Fahrer — about two years, in Derby, on a codebase they built themselves. What came out was a character recognisable to people who have never touched a joypad, a game that sold in the millions, and a studio that would be dismantled by its own success inside seven years.

The Core Design story is the clearest case study the British industry has produced in what happens when a publisher discovers an annual franchise. It’s worth telling carefully, because the easy version — plucky studio ruined by greedy suits — is true enough to be boring and wrong in the details that matter.

Derby, and the Gremlin diaspora

Core formed in 1988 when Jeremy Heath-Smith and colleagues left Gremlin Graphics in Sheffield and set up in Derby. The early catalogue is honest, unglamorous 8- and 16-bit work.

Rick Dangerous (1989) is the one worth loading. It’s an Indiana Jones pastiche — hat, whip, boulder, temple — built around a genuinely hostile idea: the traps are unfair on first contact, always, and the game expects you to die learning them and then execute a memorised route. That’s trial-and-error design, which is mostly a term of abuse, and Rick Dangerous is about the best argument for it anyone made. You’re building a solution in your head across deaths.

Chuck Rock (1991) is a caveman platformer with a belly-flop attack, arriving right as the mascot platformer boom was reshaping the industry. Thunderhawk (1993) is a well-regarded helicopter sim. Core were a competent mid-tier studio with a wide range and no signature. Then Eidos bought CentreGold, Core’s parent, in 1996.

The room where Lara happened

The origin details are documented and they’re better than the myth.

Toby Gard designed the character, and the well-worn story is that she started as a man, then as a woman with a distinctly Lara Croft-ish look under the name Laura Cruz, before the publisher pushed for something more marketable across territories. The polygon budget shaped her as much as any design intent — she’s built from a few hundred triangles, and the distinctive silhouette exists partly because a low-poly figure needs strong shapes to read at all.

The reason Tomb Raider works, though, sits somewhere the character design gets no credit for: a decision the team made about the camera and the grid.

Tomb Raider moves on a hidden grid. Lara’s movements are quantised — a standing jump covers a known distance, a running jump covers a known distance, and the level geometry is built in units that match. The tank controls that everyone complains about are a direct consequence: you rotate, you commit, you move, and the game is legible because your body’s relationship to the world is arithmetic rather than analogue.

This is the same argument I’ve made about Resident Evil’s tank controls being the point, and Tomb Raider is the stronger case, because Core built an entire platform game on it. A 1996 3D platformer with free analogue movement would have been a nightmare of missed ledges and unreadable depth. Core removed the ambiguity by removing the freedom, and the result is a game where a difficult jump is a puzzle you can solve rather than an execution you can flub.

The other thing they got right was emptiness. The tombs are big, quiet and mostly unoccupied. You spend long stretches with the sound of footsteps and water, working out where the room wants you to go. Very few games since have had the nerve to leave that much silence in, and it’s what makes the wolf, when it comes, land.

The level design earns that silence by being genuinely spatial. Core built the tombs as interlocking three-dimensional volumes — you drain a cistern on one floor and open a passage two floors down, and the game trusts you to remember the geometry rather than marking it. A Tomb Raider level is a machine you’re inside of, and the pleasure of it is close to what Dark Souls would later get credit for with its folded shortcuts. Core got there in 1996, on a PlayStation, with no map screen at all.

The machine starts

Tomb Raider shipped in 1996 and sold enormously, and Eidos — a publicly listed company whose share price now depended on one character — did the obvious thing.

Tomb Raider II, 1997. Tomb Raider III, 1998. The Last Revelation, 1999. Chronicles, 2000. The Angel of Darkness, 2003. Five sequels in seven years from a team that had taken two years on the first, while the same studio was also expected to make other games.

Toby Gard left in 1997, during Tomb Raider II’s development, over disagreements about the direction the character was being taken commercially. The designer of the most famous woman in games walked out of the building a year after she appeared. He founded Confounding Factor, made Galleon, and eventually — years later — worked on Crystal Dynamics’ Tomb Raider reboot, which is a satisfying enough ending that it almost distracts from the middle.

The sequels get harder, and that’s the tell. Tomb Raider II adds guns and people to shoot, III adds difficulty that borders on contempt, and each one refills the same shape with more of everything, because there was no time to ask a design question. This is what annualisation does. It converts a studio’s design capacity into throughput.

By The Last Revelation the team were exhausted enough to kill Lara off in the ending — a transparent bid for a break, and Eidos published Chronicles the following year anyway, built substantially from flashbacks and offcuts. The Last Revelation is the best of the sequels for exactly the reason you’d expect: Egypt gave the team a single coherent setting to build a long interlocking run of levels in, and the design briefly gets room to breathe again.

Angel of Darkness

The Angel of Darkness (2003) is one of the most instructive failures in the medium, and the reason to be precise about it is that it was ambitious.

Core wanted a different game: a darker one, with conversation choices, stealth, an RPG-ish strength system where Lara got physically stronger over the story, and a second playable character. That’s a real design idea. It’s also a wholesale reinvention, attempted on a new engine, for the PS2, by a team that had been shipping annually for six years.

It shipped broken. Missing features, incoherent controls, a story that stops mid-arc. The reviews were brutal and the sales weren’t, and Eidos moved the series to Crystal Dynamics in America. Two planned sequels died with it.

The honest reading: Angel of Darkness is what an exhausted studio’s ambition looks like when it finally gets a chance and gets it at the worst possible moment. Heath-Smith resigned in the aftermath.

The long fade

Core spent its remaining years on work-for-hire. A remake of the original Tomb Raider for the tenth anniversary was well under way — a project the team had considerable enthusiasm for — and Eidos cancelled it in favour of Crystal Dynamics’ Anniversary. Rebellion bought Core in 2006 and renamed it Rebellion Derby. It closed in 2010.

That arc — bought by a publisher, milked, discarded, renamed, shut — is the standard shape, and I’ve traced the same one through Bullfrog’s absorption into EA and DMA’s passage through three owners. The 90s British studios almost all end this way. Core’s version is the harshest because the asset they created was so unusually valuable that nobody was ever going to let them rest it.

What Core actually proved

Two things worth keeping.

The first is a design principle with a long life: constraint produces legibility. Tomb Raider’s grid, its quantised jumps and its refusal to give you a free camera are all restrictions, and they’re precisely what makes a 1996 3D space navigable. The modern games that inherited from it — every ledge-grabbing, wall-running descendant — solved the same problem by making the character stick to everything automatically, which is a solution that removes the puzzle along with the difficulty.

The second is about labour, and it’s the one the industry keeps relearning. A studio’s output is design capacity, and design capacity does not scale by scheduling. Core in 1996 had two years and six people and produced a landmark. Core in 2003 had a bigger team, a bigger budget, seven years of pipeline and no room to think, and shipped a wreck. The input that ran out wasn’t money.

Where to start

The original Tomb Raider, on PC or PS1, and give it an hour before you judge the controls. The grid is a language and it takes about forty minutes to learn.

Rick Dangerous on the Amiga, for the studio’s first and purest statement about learning by dying.

The Angel of Darkness only if you want to see what the studio was trying to become. It’s a sad document and it’s an honest one.