By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

How AI Could Bring Halo and Other Classic Games Back to Life

AI is collapsing the distance between an idea and a playable game. What happens when almost anyone can build one?

Jonathan LaiguestAnish AcharyaguestErik Torenberghost

Nov 12, 202646mWatch on YouTube

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

Great conversations run long. Their best ideas don't have to.

uListen is free on iPhone, iPad, Android, Chrome, and Claude.

CHAPTERS

Jump to publisher timestamps

Chapter summaries

Why AAA games take a decade—and why AI could change the pace
The conversation opens on the extreme length and cost of modern game development, using Grand Theft Auto VI as an example of a 12–13 year cycle. The speakers frame AI as a potential inflection point that could make game creation faster and accessible to more people.
- •Grand Theft Auto VI is cited as an example of a decade-plus development cycle
- •Games are described as uniquely complex, multidisciplinary products to build
- •The group tees up AI as a shift that could reshape how games get made and who can make them
Astra’s key unlock: agents that can directly operate game-dev tools
Jonathan describes a recent “technology unlock” where Astra can use popular game development software directly instead of only giving instructions. He walks through an end-to-end flow from design doc to playable build, including AI-generated assets and sound.
- •Astra is described as being able to operate Blender, Unreal, and Unity directly
- •A full pipeline is outlined: design doc → concept art → 3D models/rigging/animation → engine integration → audio → export
- •Astra is described as enabling “one-shot” playable prototypes from a single workflow
Self-QA and the shrinking gap between idea and prototype
Jonathan explains that Astra can QA its own output by inspecting screenshots and iterating on bugs and visual issues. The result, they argue, is surprisingly strong prototypes and a dramatic reduction in time from concept to something testable.
- •Astra is described as being able to QA itself using screencaps and feedback loops
- •Shorter iteration cycles are positioned as the first major industry ramification
- •Fast prototyping is framed as enabling experimentation with “strange or weird” ideas
What shorter dev cycles unlock: niche genres, more experiments, bigger ambition
The speakers argue that long dev cycles suppress innovation because producers become conservative, favoring sequels and proven genres. AI-driven speedups could enable more new IP, more creative risk, and even higher ambition at the top end.
- •Long timelines are linked to canceled projects and creative ideas going stale
- •Conservatism in green-lighting is tied to sequels and established shooter genres
- •Shorter cycles are framed as enabling more experimentation and more diverse games
- •Anish suggests reduced friction could also increase ambition—expanding scope from cities to “universes”
AI changes who can make games: lowering the multidisciplinary barrier
Jonathan emphasizes that making games historically requires coding, art, design, and audio expertise, creating a high execution floor. With AI filling skill gaps, he argues hobbyists can reach prototypes and some will become sustained creators.
- •Game creation is described as a true multidisciplinary craft with a high execution floor
- •AI is framed as filling specialized knowledge gaps so creators don’t need full teams
- •Jonathan predicts more people will prototype, with some continuing into serious development
Can AI make a game fun? The last 20% and the role of human taste
Erik asks whether AI can handle balance and “fun,” and Jonathan argues human taste and judgment remain central. Jonathan describes the industry saying that the last 20% of a game is 80% of the work, driven by feel, detail, and player desire.
- •Human taste and creative judgment are described as not replaceable by models (for now)
- •The ‘last 20% is 80% of the work’ framing is used to explain what makes games fun
- •Examples of “feel” include sound design, weapon recoil, and experiential details
- •Erik highlights depth, consequence, and immersion as what drives long playtime
When creating becomes the game: personal, tranquil game-making as play
Anish proposes that in the future many games will be played primarily by their creators—and that’s fine—because creation itself can be satisfying. Jonathan connects this to Minecraft and Legos, arguing AI turbocharges the joy of bringing worlds to life.
- •Anish argues “creation is consumption,” with many games primarily experienced by their makers
- •Game-making is described as tranquil, with a “zen gardening” aspect
- •Jonathan links the idea to Minecraft and the broader concept that creation itself is play
- •AI is framed as enabling people to build things they can feel proud of without mastering every discipline
Case study: building a Habbo Hotel-inspired online record shop—then a whole street
Anish describes building a Habbo Hotel-themed record shop inspired by the social experience of record stores. He extends it into a 1990s Toronto street where other players can create new storefronts via prompting, including new mechanics like adoptable pets that follow you.
- •The project begins from a critique that online record stores are just lists, lacking social atmosphere
- •Using Opus 5.5, the system generates a room and unexpected genre-specific music crates
- •Anish expands the scene into a full street meant for other players to build via prompts
- •A coding agent is described as updating the game engine to add new mechanics (pets that follow you)
- •They contrast procedural content generation with “dynamic abilities” and new mechanics
His son rebuilt Ghost of Tsushima in Minecraft—and what remixing means
Jonathan shares how his eight-year-old used GPT and existing wikis to deconstruct Ghost of Tsushima’s mechanics and recreate it as a Minecraft mod after being told he couldn’t play the original. The story becomes a springboard for the idea that remixing concepts can matter more than pure realism.
- •Jonathan’s son recreates Ghost of Tsushima as a Minecraft mod using GPT/Codex and online references
- •Jonathan notes the mod works well and is shared with friends, reducing the need to buy new games
- •Anish argues realism isn’t the only important axis; combining concepts can be the breakthrough
- •The conversation transitions from modding into the broader trend of decompilation and recombination
Mods, decompilation, and reviving classics: from Counter-Strike to Halo on the web
Jonathan and Anish outline AI-turbocharged paths: modding as a historical engine of major franchises, decompiling old games to regain source/assets, and combining assets from different games. Jonathan cites community projects that decompiled and recompiled the original Halo into playable web GPU versions, connecting this to platform-holder interest.
- •Modding is described as foundational to major franchises (e.g., Counter-Strike, Dota/LoL inspirations, battle royale mod lineage)
- •AI is framed as making it easier to remix existing games rather than start from a blank canvas
- •Anish highlights decompiling to recover source/assets for beloved older games and extend or fork them
- •Anish describes combining assets from two games into a third as a futuristic approach
- •Jonathan cites multiple community efforts making the original Halo playable in the browser via web GPU
- •AI is framed as helping with porting/remastering to reach new audiences on modern devices
The renaissance of niche genres: RTS hopes and job-simulator obsession
Anish argues high-cost genres with unclear ROI—like RTS—could see a revival with AI lowering production barriers. Jonathan riffs on niche “job simulators,” sharing his enthusiasm for a train conductor simulator, and the group reflects on how simulator-style play has long appealed to players.
- •RTS is described as beloved but historically difficult to justify economically, especially on new platforms like iPad
- •Anish predicts a renaissance in RTS as costs fall and feasibility rises
- •Jonathan highlights niche job simulators and describes a detailed train conductor simulator loop
- •They discuss how simulator genres have enduring appeal, including references to flight sims and Untitled Goose Game
- •The segment frames AI as enabling more ultra-specific experiences for smaller audiences
Managing AI agents like an RTS: why game-like UIs may win
Erik notes emerging agent orchestration interfaces skinned like classic RTS games, asking if it’s just a toy or a new genre. Anish compares agent management to running a business with limited control and imperfect information, while Jonathan argues consumers need more intuitive visual interfaces than command lines.
- •RTS-style skins for agent orchestration are discussed as an emerging pattern
- •Anish likens managing agents to managing teams and market forces in a business sim
- •The group argues falling UI/implementation costs make ambitious interfaces more viable
- •Jonathan says command-line agent management is non-intuitive for average consumers
- •They expect experimentation in “agentic workflow” interfaces, whether or not it looks like StarCraft/Age of Empires/Factorio
The business of AI games: who captures value, and what distribution might change
Anish asks whether major ecosystems are positioned to capture value from AI-created games and modding. Jonathan argues IP owners historically haven’t benefited much from mods, raising questions about revenue share, royalties, and new distribution platforms built for AI-native creation. Anish points to Spawn as an early example of a shared community plus game-generation platform.
- •Jonathan says IP owners often didn’t capture much value from mods built on top of their games
- •They discuss the difference between extending a base game (benefiting the owner) vs forking (splitting community)
- •Jonathan argues there’s room for AI-native distribution platforms beyond Steam/App Store/Roblox channels
- •Anish describes Spawn: an embodied agent that can take bug reports, modify games while you’re in them, and create portals between multiplayer experiences
- •They compare Minecraft’s rich shared core with Roblox’s more decentralized entry points and experiences
What excites and scares studios—and the ‘never-ending game’ vision
Erik asks what publishers and studios are most excited and worried about. Anish emphasizes cost reduction, lower risk, and improved quality of life, while also highlighting the desire for a “forever game.” Jonathan frames current AI use as wave one (productivity) and wave two as new mechanics and genres, then describes the ‘never-ending game’ idea: an MMO that adaptively generates content based on what players actually do.
- •Studios are said to be excited by lower costs, lower financial risk, and the ability to create more IP
- •Anish describes harsh studio dynamics and hopes AI improves industry quality of life
- •The “forever game” is framed as a hard-to-achieve holy grail with persistent relevance
- •Jonathan describes wave one AI adoption as productivity (design docs, concept art, prototyping)
- •Jonathan describes wave two as AI enabling new mechanics/genres rather than just faster versions of existing games
- •The never-ending MMO concept: the game adapts quest lines and content to player choices instead of linear scripting
The games they’d finally build: recreating lost classics, agent-driven Civ, and AI dating sims
To close, the speakers share the games they’ve put the most time into and what they’d build now with AI. Anish describes recreating a lost-source-code BBS game from an EXE and design doc; Jonathan wants to modernize Civ II and imagines a strategy game where units behave like agents. Erik shares his own Unity experimentation, leading into Jonathan’s argument that dating sims are especially ripe for AI-driven characters.
- •Anish cites Witcher 3 and Dune 2000 as major personal favorites, including a port to Mac for Dune 2000
- •Jonathan cites League of Legends and StarCraft, and discusses RTS difficulty and market constraints
- •Anish describes recreating Solar Realms Elite despite lost source code, using only a design doc and an EXE
- •Jonathan wants to recreate Civ II for modern hardware and imagines Civ-style play with agent-like units and agendas
- •Erik recounts building a basic dating-sim prototype in Unity during the pandemic
- •Jonathan argues otome/dating sims could shift to agent-driven characters with personalities, lore, and real-time interaction

Publisher timestamps

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.