The Next Era of Games: Mods, Agents, and Infinite Worlds

Original Media · 46m · English · Podcast · The a16z Show

TLDR

A16Z General Partners Anish Acharya and Jonathan Lai discuss how AI tools like Astra — which can directly operate Blender, Unreal, and Unity — are collapsing game development cycles from years to days, enabling solo creators to prototype, mod, and even decompile classic games. They explore how cheaper, faster production could unlock niche genres, revive dead IP, and shift the industry from risk-averse sequels toward experimental new IP, while raising questions about distribution, business models, and the future of adaptive, "never-ending" games that generate content in real time.

For you

PRO members can sign in and request more information — or their own version of this digest, tuned to what they care about.

Sign in to unlock

Surprising Insights

•

An eight-year-old with no coding background shipped a working Ghost of Tsushima Minecraft mod by prompting Codex to read wikis and generate the mod code.

•

The host recreated a lost 1990s BBS game (Solar Realms Elite) from only an EXE and a design doc — something every prior model generation failed to do.

•

Crusader Kings 3 already implements the "ruler with agent subjects" fantasy Jonathan imagined for Civilization, without frontier AI — suggesting some AI-native designs may arrive via traditional systems first.

Key takeaways

•

AI models like Astra can now directly manipulate professional game-dev tools (Blender, Unreal, Unity) rather than just explaining how to use them, letting a complete novice draft a design doc, generate concept art, build 3D models, animate, rig, import to an engine, add audio, and QA the result in a single session.

•

Development cycles that historically took 10–13 years (e.g., Grand Theft Auto 6) could shrink to 2–5 years, lowering the financial risk of greenlighting new IP and weird genres — like job simulators or esoteric RTS games — that previously couldn't justify the investment.

•

Modding, the origin of billion-dollar franchises such as Counter-Strike, League of Legends, and Fortnite, is being supercharged: players can point an agent at a game and say "make this but add voxel crafting from Animal Crossing," instantly creating playable hybrids.

•

Decompilation and recompilation of classic titles (Halo 2001 in WebGPU, Knights of the Old Republic) are letting communities port abandoned IP to modern platforms without publisher involvement, exposing legacy games to new audiences.

•

The skill floor for game creation is disappearing: an eight-year-old built a Ghost of Tsushima Minecraft mod by asking Codex to deconstruct wiki mechanics, and the host's son now creates games instead of asking to buy them.

•

Studios are already using AI for design docs, concept art, and rapid prototyping (wave one); the next wave is AI-native gameplay — agents as NPCs with their own agendas, dynamically generated quest lines, and worlds that adapt to player choices in real time.

•

Current distribution (Steam, App Store, Roblox) isn't built for the flood of AI-generated games; new platforms like Spawn — where an embodied agent co-pilots creation, modifies games live, and links worlds via persistent community — hint at a native distribution layer.

•

Publishers and platform holders (Epic, Xbox) recognize the shift but face a business-model gap: historically they captured little value from mods (Half-Life didn't profit from Counter-Strike), and they now need revenue-share or royalty mechanisms for AI-remixed UGC.

Notable quotes

“Grand Theft Auto 6 is finally coming out next month. That game has been in production since like 2013, 2014. It's like a 12, 13-year development cycle.”

▸ Notable quotes · 0:00

“Astra was very good at using Blender. It could actually work directly within Unreal and Unity. And in previous generations of models, they would just explain to the user, oh, you should do this in Blender, you should do this in Unreal. But Astro actually could like operate the software for you.”

▸ Notable quotes · 2:25

“The last 20% of the game is 80% of the work. And that's where all the magic comes together that makes something fun.”

▸ Notable quotes · 10:45

“Creation is consumption. Like, the vast majority of games are ever going to be played by one person. That is the creator in the future. And that's okay, actually.”

▸ Notable quotes · 12:27

“In an adaptive MMO, one that, you know, maybe is built using like world models, for example, like you could literally generate an entire quest line for someone who really likes the ocean.”

▸ Notable quotes · 36:10

“Historically the original owners of IP have not benefited that much from the mods that have been created on top of their IP, right? And so I think the original creators of Dota didn't really benefit that much from League of Legends when that was created.”

▸ Notable quotes · 28:11

Chapters

•

2:18 — AI tool-use breakthrough with Astra: Anish describes how Astra can directly operate Blender, Unreal, and Unity — drafting design docs, generating art, building 3D assets, animating, rigging, importing, adding audio, and self-QAing — enabling a hobbyist to one-shot a playable prototype.

•

3:35 — Shrinking dev cycles and unlocking niche genres: Long cycles (GTA 6's 13 years) force conservative greenlighting; cutting to 2–5 years lets studios take creative risks on new IP and esoteric genres like job simulators that were previously economically impossible.

•

18:11 — Modding, decompilation, and remixing: AI supercharges the modding lineage (Counter-Strike, LoL, Fortnite) — players can remix games by prompting agents; communities are decompiling classics like Halo and Knights of the Old Republic for modern platforms without publisher help.

•

8:40 — Democratizing who can make games: Game creation's multidisciplinary barrier (code, art, design, audio) collapses as AI fills skill gaps; an eight-year-old built a Ghost of Tsushima Minecraft mod, and the host's son now creates games instead of asking to buy them.

•

36:00 — AI-native gameplay and the never-ending game: Beyond faster production, AI enables adaptive worlds: agents as NPCs with agendas, dynamically generated quest lines, and games that reshape themselves to player choices — the "mind game" from Ender's Game.

•

29:55 — Distribution and business-model gaps: Traditional stores (Steam, App Store) aren't built for AI-generated volume; platforms like Spawn experiment with persistent community and live agent co-piloting; publishers need new royalty models since they historically captured little mod value.

•

39:54 — Personal dream projects: Anish recreated a lost 1990s BBS game from an EXE and design doc; Jonathan wants to revive 16-bit Civilization with agent-driven citizens; the host prototyped a Japanese dating sim in Unity during COVID.

Speakers

•

Anish Acharya — A16Z General Partner

•

Jonathan Lai — A16Z General Partner

•

Host — Podcast host

⋮