A 102M-parameter probabilistic forecaster called t0-alpha, released in June 2026 by The Forecasting Company, demonstrates the recipe behind many current time-series LLMs. The model cuts a numerical sequence into patches of 32 time steps, processes them with a causal transformer, and emits quantiles rather than a single future line. On GIFT-Eval, t0-alpha reproduced its reported headline numbers exactly: CRPS 0.4941 and MASE 0.7240. The model is small enough to run on accessible hardware, and its weights are published openly under Apache-2.0. The post uses t0-alpha to explain how time-series foundation models work, how they are evaluated, where they beat classical baselines, where they still fail, and why the next useful gains may come from calibration, routing, leakage control, stronger baselines, and domain-specific estimators.