D3D introduces a standalone diffusion drafter for asynchronous speculative decoding that eliminates inter-round dependencies between drafting and verification. By combining blockwise denoising with auxiliary autoregressive supervision, D3D achieves long accepted prefixes without requiring updated target-model hidden states for each new block. This decoupling enables parallel token prediction while maintaining high acceptance rates, overcoming the sequential bottleneck in prior target-conditioned diffusion drafters.