Andrej Karpathy Tests AI With Lord of the Rings 3D World Project Using Claude Opus 5
AI researcher Andrej Karpathy tested Claude Opus 5 by having it generate a 3D interactive world from the opening of The Lord of the Rings. The experiment showcased advanced multi-hour coding capabilities but also revealed ongoing limitations in real-time visual self-auditing.
Artificial intelligence researcher Andrej Karpathy has demonstrated a significant shift in how large language models are evaluated, using a creative test involving J.R.R. Tolkien’s The Lord of the Rings. Moving away from traditional, simple prompt benchmarks, the experiment highlighted both the expanding capabilities and the remaining architectural limits of frontier models.
Lord of the Rings 3D World Test Using Claude Opus 5
Karpathy provided the Claude Opus 5 model with the opening paragraph of The Lord of the Rings, allocating a one million token budget valued at roughly USD 10. He instructed the system to render the literary scene as an interactive three-dimensional web environment using Three.js. Operating autonomously for approximately two hours, the model produced around 5,500 lines of JavaScript code. The resulting output procedurally generated a low-poly representation of the Shire, complete with environmental assets, character structures, and synchronized animations.
The experiment illustrates how artificial intelligence can tackle hyper-custom tasks that humans would rarely spend time executing manually due to labor constraints. Because language models possess vast computational patience, bespoke digital environments can be generated at a minimal marginal cost. This capability opens new possibilities for dynamic, on-demand content creation where virtual worlds or specific software simulations can be materialized instantly through simple narrative descriptions.
Despite the successful code generation, the test exposed critical gaps in multimodal reasoning and self-correction. Because current large language models lack native, real-time video perception and the ability to natively play or test interactive environments, the model struggled to evaluate its own visual results. Instead of seamlessly experiencing the rendered space, the model had to rely on slowly capturing individual screenshots, which led to various graphical flaws and structural inconsistencies.
(The above story first appeared on LatestLY on Aug 03, 2026 06:46 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).
