Definition
Plain language
A technique that automatically packs more detail into the parts of a simulation where things are changing fast, and less where they're calm.
As stated in the literature
A numerical method that dynamically refines the computational grid in regions of high gradient or error, concentrating resolution where the solution varies sharply; can conflict with positivity-preserving limiters, forcing pre-graded meshes.
Also called: AMR
Why it matters: It lets simulations capture sharp, fast-changing details without wasting compute on the calm regions, making otherwise unaffordable problems tractable.
For example, in a simulation of a breaking wave, the method automatically uses a fine grid right at the crest where the water churns and a coarse grid in the still water far away.
Heard on the show
“Then it notices that limiter blocks adaptive mesh refinement, so it pre-grades the mesh in tight rings around the bow shock to compensate.”Episode 042 — An Agentic Scientific Computing System That Actually Remembers What It Learns