This paper presents a new pretraining method, CellWorld, for foundation models in spatial transcriptomics. Existing spatial transcriptomics foundation models primarily reconstruct masked gene identities or expression values, potentially encouraging the reproduction of assay-specific technical variation and limiting representation transferability. To avoid directly reconstructing such variation, we shift the prediction target from observed gene measurements to latent cell representations. CellWorld predicts the latent representations of masked cells from visible spatial context and a limited partial-expression hint. We pretrain four CellWorld variants, spanning 5.74M to 94.56M trainable parameters, on a corpus of 46 million human cells. Our controlled scaling experiments show that performance improves with model capacity, particularly on spatial tasks, while spatial transfer depends more on sufficient optimization and broad biological source diversity than on cell count alone. Across four held-out datasets, even CellWorld-Small, with 5.74M trainable parameters, outperforms every baseline on all 11 linear-probe benchmarks and all seven fine-tuned spatial benchmarks. Most notably, a frozen CellWorld-Large pretrained on only 5% of the corpus with broad biological source coverage outperforms every fully fine-tuned baseline across all seven spatial benchmarks. Code is available at https://github.com/UoM-HealthAI/CellWorld. Blogger's Review: This paper proposes a new pretraining method, CellWorld, for spatial transcriptomics foundation models, which predicts latent cell representations to avoid direct reconstruction of technical variation. The experimental results show that CellWorld outperforms existing baseline models and has a wide range of application prospects.