--- schema: 1 type: video title: 'The Next Game Engine Won''t Have a Manual — Arturo Nunez, Nereu' description: >- Ask a coding agent for a camera that follows your character and it will reinvent that camera from scratch, every time, slightly differently. Arturo Nunez's diagnosis is that the context sits on the game engine's vocabulary rather than the game's. Controlling a character in a conventional engine means a mesh, a renderer, an animator, a rigid body, a collider, an audio source, and only then your actual movement logic, nearly all of which is boilerplate that every character in every game already carries. Nereu inverts that. Everything is an asset, and you attach tags describing intent instead of implementation: character, animated, double jump. Systems then query by tag and move everything marked vehicle and drivable, which is Entity Component System thinking lifted from data oriented design. The pleasant consequence is that nothing stops you tagging a building as drivable and dropping it into a Mario Kart style race. The assistant is there to get you unstuck rather than to one shot a finished game, and the vocabulary it expects is the one tutorials already use: press A to jump, press A again in the air. The engineering detail worth stealing is how context gets assembled. Rather than feed the whole scene to a model, he borrows level of detail from rendering. Assets near whatever you are editing arrive with their full tag values, distant ones collapse to a position and a type, and the hundred pieces of grass are simply left out. Speaker info: - https://x.com/arturonereu - https://www.linkedin.com/in/arturonereu/ - https://www.arturonereu.com/ Timestamps: 0:00 - Building a game live by describing it 2:45 - Why making games is hard 4:27 - Ten years at Unity watching the same struggles repeat 6:59 - Powerful engines and LLMs that still do not compose 7:49 - The boilerplate behind controlling a character 8:45 - Everything is an asset, and tags describe intent 9:37 - The asset tag system, and tagging a building as drivable 11:21 - How the prompt gets its context 14:52 - Level of detail, applied to context assembly 16:37 - Getting unstuck rather than one shotting a game 17:28 - World models are a different medium canonical: 'https://liquidgold.cc/watch/VBCDhRrvlYo' markdown: 'https://liquidgold.cc/watch/VBCDhRrvlYo.md' llms: 'https://liquidgold.cc/watch/VBCDhRrvlYo/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T05:13:38.431Z' language: null children: [] videoId: VBCDhRrvlYo durationSeconds: 1173 publishedAt: '2026-08-18T15:00:29.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 22 endSeconds: 176 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=22' heldReason: rights - ordinal: 1 text: null startSeconds: 22 endSeconds: 176 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=22' heldReason: rights - ordinal: 2 text: null startSeconds: 176 endSeconds: 186 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=176' heldReason: rights - ordinal: 3 text: null startSeconds: 186 endSeconds: 276 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=186' heldReason: rights - ordinal: 4 text: null startSeconds: 287 endSeconds: 425 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=287' heldReason: rights - ordinal: 5 text: null startSeconds: 436 endSeconds: 470 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=436' heldReason: rights - ordinal: 6 text: null startSeconds: 470 endSeconds: 480 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=470' heldReason: rights - ordinal: 7 text: null startSeconds: 525 endSeconds: 536 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=525' heldReason: rights - ordinal: 8 text: null startSeconds: 580 endSeconds: 593 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=580' heldReason: rights - ordinal: 9 text: null startSeconds: 580 endSeconds: 593 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=580' heldReason: rights - ordinal: 10 text: null startSeconds: 682 endSeconds: 692 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=682' heldReason: rights - ordinal: 11 text: null startSeconds: 692 endSeconds: 894 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=692' heldReason: rights - ordinal: 12 text: null startSeconds: 894 endSeconds: 904 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=894' heldReason: rights - ordinal: 13 text: null startSeconds: 1004 endSeconds: 1015 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=1004' heldReason: rights - ordinal: 14 text: null startSeconds: 1050 endSeconds: 1062 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=1050' heldReason: rights - ordinal: 15 text: null startSeconds: 1062 endSeconds: 1173 watch: 'https://liquidgold.cc/watch/VBCDhRrvlYo?t=1062' heldReason: rights --- # The Next Game Engine Won't Have a Manual — Arturo Nunez, Nereu Ask a coding agent for a camera that follows your character and it will reinvent that camera from scratch, every time, slightly differently. Arturo Nunez's diagnosis is that the context sits on the game engine's vocabulary rather than the game's. Controlling a character in a conventional engine means a mesh, a renderer, an animator, a rigid body, a collider, an audio source, and only then your actual movement logic, nearly all of which is boilerplate that every character in every game already carries. Nereu inverts that. Everything is an asset, and you attach tags describing intent instead of implementation: character, animated, double jump. Systems then query by tag and move everything marked vehicle and drivable, which is Entity Component System thinking lifted from data oriented design. The pleasant consequence is that nothing stops you tagging a building as drivable and dropping it into a Mario Kart style race. The assistant is there to get you unstuck rather than to one shot a finished game, and the vocabulary it expects is the one tutorials already use: press A to jump, press A again in the air. The engineering detail worth stealing is how context gets assembled. Rather than feed the whole scene to a model, he borrows level of detail from rendering. Assets near whatever you are editing arrive with their full tag values, distant ones collapse to a position and a type, and the hundred pieces of grass are simply left out. Speaker info: - https://x.com/arturonereu - https://www.linkedin.com/in/arturonereu/ - https://www.arturonereu.com/ Timestamps: 0:00 - Building a game live by describing it 2:45 - Why making games is hard 4:27 - Ten years at Unity watching the same struggles repeat 6:59 - Powerful engines and LLMs that still do not compose 7:49 - The boilerplate behind controlling a character 8:45 - Everything is an asset, and tags describe intent 9:37 - The asset tag system, and tagging a building as drivable 11:21 - How the prompt gets its context 14:52 - Level of detail, applied to context assembly 16:37 - Getting unstuck rather than one shotting a game 17:28 - World models are a different medium Canonical: https://liquidgold.cc/watch/VBCDhRrvlYo · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.