It’s worth investing heavily in Three.js and WebGPU now. Not only because of the imminent visual power of AR/VR, but also because of a critical reality: AI democratizes bloat code, and hardware penalizes inefficiency.

I think the next big market niche will be for engineers who understand the true cost of computing for webgpu computational cost and can audit and debug the chaos generated by even the simplest warning.

The Inevitability of the Graphics Paradigm and the AI ​​Entropy Crisis: Considering whether learning Three.js and WebGPU is ‘worth it’ is a category mistake; it’s not a choice, it’s a technical survival imperative. We are witnessing the obsolescence of DOM-based rendering in favor of hardware-accelerated spatial computing. However, the real investment thesis is not visual fidelity, but the entropy of AI-generated code. LLMs are probabilistic engines that generate syntactic ‘slop’: code that compiles but is algorithmically inefficient and structurally redundant. AI can hallucinate logic, but it cannot hallucinate physics. The GPU is a strictly deterministic environment; ‘implementers’ will be replaced by AI. The real value will be concentrated exclusively in Graphics Optimization Engineers: those capable of auditing AI’s probabilistic garbage, refactoring it at a low level (WebGPU/WGSL), and forcing its Execution within the 16ms frame budget limits. We’re not learning graphics to ‘create’ better, but to prevent the web from drowning in its own automatically generated inefficiency.