Artificial intelligence could help game developers attempt more ambitious projects by filling gaps in their skills and shortening the path to working prototypes, venture capital firm Andreessen Horowitz said.
In “Gaming’s Next Origin Stories,” published Nov. 12, the firm argued that AI could strengthen independent developers, technical pioneers and large studios. Its analysis drew on personal projects and industry examples rather than a study covering a defined period.
The authors described opportunities both to reduce the cost of making games and to create experiences that developers previously could not attempt. They considered work on new games alongside changes to existing titles, including rebuilding older software.
For independent developers, the immediate benefit could come from filling gaps between an idea and the skills needed to build it, the authors said. They described AI agents helping write design documents, create visual assets, assemble scenes, inspect screenshots and attempt repairs.
A working prototype still leaves considerable work before a finished game, the authors said. As an example of experimentation, they described contributor Anish A building a virtual record shop inspired by shopping in Toronto in the 1990s.
The authors said Anish later allowed players to change the shop and its surrounding street. When his son added a pet-adoption center, a coding agent added animals that followed their owners and adapted the original software to support the new behavior.
AI could also help recover older games through decompilation, the authors said, describing the process of turning compiled software back into code people can inspect and change. They cited a project led by Maurice Heumann in which agents initially produced readable code that behaved incorrectly.
Heumann’s team responded by checking reconstructed functions against the original program, the authors said. They presented that work as an example of engineers designing checks that allow AI agents to tackle difficult tasks more reliably.
For large studios, shorter development cycles could make unfamiliar ideas easier to attempt, the authors argued. Committing years and hundreds of millions of dollars to a game encourages producers to favor familiar franchises, while faster prototypes could allow more experiments.
The authors also described efforts to recreate games whose original code was unavailable. Anish used a surviving design document and executable program to rebuild Solar Realms Elite, returning to the project as newer models became available.
The authors acknowledged criticism that early AI-made games may entertain their creators and small groups of friends without offering broader appeal. They also acknowledged questions about whether the enjoyment comes from making a game rather than playing it.
They nevertheless argued that creating games can itself be play and that experimentation can produce ideas others develop further. They cited Counter-Strike’s beginnings as a modification of Half-Life in 1999, while saying such experiments need not produce another comparable success.
