The Bank for International Settlements has published a working paper introducing the BIS Time-series Regression Oracle (BISTRO), a general-purpose “foundational” model designed to produce both unconditional (baseline) and conditional (scenario) forecasts for macroeconomic time series without requiring task-specific model redesign. BISTRO adapts transformer methods used in large language models and is fine-tuned on a BIS macroeconomic dataset covering 4,925 time series across 63 economies, with the training and evaluation window set at 1984–2024. The paper provides operating scripts and a guided workflow that can be run in Google Colab, allowing users to upload their own data and generate probabilistic baseline and conditional forecasts while accounting for mixed frequencies and publication lags through daily alignment. In out-of-sample comparisons against an AR(1) benchmark and the pre-trained MOIRAI model across multiple evaluation windows and horizons, BISTRO generally performs better for unemployment and GDP growth and improves relative to benchmarks at longer inflation horizons, and it is illustrated as supporting conditional scenario analysis such as alternative oil price paths that produce non-linear inflation responses.