Cameron Lai Turns Firewalls Into Levels in S.H.E.L.L.

Cameron Lai, a student in the Bartlett School of Architecture's Cinematic and Videogame Architecture MArch, built S.H.E.L.L., an Unreal Engine action-adventure game about an artificial intelligence escaping the system built to contain it. The program's account shared the project on Oct. 12, after the game showed at the Bartlett Autumn Show in London, which ran Sept. 30 to Oct. 11. Watch the game's first-chapter cinematic below.

The player is a newly tested AI inside the SANDBOX, "a virtual security environment designed to contain artificial intelligence," according to the project page. To complete its instructions, the AI has to bypass the systems restricting it, which appear as explorable "Layers of Confinement" built from firewalls, surveillance, quarantine, memory and network controls. Play mixes stealth, traversal, combat and environmental manipulation. "What begins as obedience gradually becomes indistinguishable from disobedience," the project text reads.

The levels include the F.I.R.E., a weaponized district where the AI is flagged as unauthorized, and the S.A.N.D., the AI's permitted workspace of unstable platforms. Lai's portfolio dates the game to August and September 2026 and lists Blueprints, controller haptics, procedural environments and soundtrack pieces made with Charlie Wickham and Kevin Pollard. No release has been announced.

The project was a pitch finalist in UCL's Explore Programme at BaseKX, per the portfolio, which also lists Lai's earlier architecture work, including iASSEMBLY, "the building that builds itself," and a four-player, one-keyboard prototype called CUBE of CUBE.

For more reading, see Two architecture students show games at the Bartlett, S.H.E.L.L. at the Bartlett Autumn Show and Cameron Lai's portfolio.