Real Neural Forge is a forecast dashboard for a time-series model. The application displays recent historical data, upcoming predictions, and comparisons with baseline models. Users can view shared forecasts, a lunar calendar, and near-Earth object panels without needing to sign in. The dashboard requires JavaScript to load its charts and visualizations. The platform includes features for evaluating forecast accuracy against baselines and comparing predictions with actual outcomes. It provides one-step predictions plotted over a recent time window. Users with an account can access a private workspace where datasets, templates, and models are stored. Training a model writes a private checkpoint, while shared starter files remain read-only. Each new account receives sample datasets and a starter model for immediate use. A background training run can be started using the current configuration, and a token saved in the browser is sent with training requests to identify the owner account on the public server. The application draws on data from sources such as JPL predicted approaches to Earth and the Moon, SATCAT satellite decay dates, and the model’s own 24-month outlook. Visitors who have not signed in can use the account forms at the top of the page to create an account or sign in. The typical workflow involves loading a sample dataset, selecting a model (either the starter or a trained version), running predictions, and comparing metrics across different runs.