A small laboratory for time & space

Pay for
forgetting.

Twelve tasks. Four cache slots. Every forgotten result has to be computed again. Can you explain where the extra work comes from?

Step through the plan

The dependency graph

TARGET: L LIVE
Cache Foundry dependency graphArrows point from required inputs to output tasks. A complete text dependency list follows the diagram.
CachedRunnable nowNext / failed action
0Computations so far
0 / 4Cache occupancy
0 / 29Actions completed

Cache slots · displayed alphabetically

Read the graph and rules
  1. Start with an empty cache. Each output occupies one slot.
  2. Run a task only when it is absent, every direct input is cached, and one separate output slot is free. Inputs stay cached.
  3. Every run costs one computation, including source tasks A, B and C. Dropping a cached task is free.
  4. No overwrite, fused run/drop, parallel actions, external storage or partial outputs. Finish with L cached; other outputs may remain.

Inspect a plan

Showing the published founder baseline.

Exactly capacity and actions. Up to 64 KiB UTF-8, 200 actions, capacity 1–12. Rules and dependencies cannot be overridden.

Action trace

    Select a row to inspect its resulting state. An illegal action shows the last legal state.

    Build a plan by hand

    Start empty, then run or drop tasks. Buttons use the plan’s final legal state, independent of the trace position.

    This is a browser-side companion, not the authoritative verifier. The source bundle includes two independent exhaustive checks and a lower-bound certificate. Open contributions include a short proof, a reproducible audit, or a clearly labelled changed-graph study.