Definition
Plain language
The nuts-and-bolts arithmetic choices inside a simulation that decide how accurate and how fast it runs.
As stated in the literature
The discretization schemes, solvers, and finite-precision arithmetic used in a computational model; tuning them trades stability, accuracy order, and wall-clock cost.
Also called: numerical scheme
Why it matters: Bad numerical choices can make a simulation blow up into nonsense or burn enormous compute for accuracy nobody needed.
For example, a weather simulation that steps forward one minute at a time is far more accurate but far slower than one stepping forward an hour at a time.
Heard on the show
“The fourth involves optimizing the numerics, inside a physics-based weather forecasting model, which has nothing to do with anything the loop ever touched.”Episode 276 — An AI Agent Rewrote Its Own Scaffolding For Eight Days. Here's What Survived