AptifoldContact

APTIFOLD / RESILIENCE DESIGNER

Which backup is worth it?

You’re the dispatcher. Book transport and storage so each depot gets its order on time. Then close a route and see what still arrives.

A fictional freight problem. Costs are reservation units, not real prices.

Go to the route test

ILLUSTRATIVE DISPATCH / Supply network

Eight crates. Two receiving windows.

Try it: close the north route. Can the same bookings still fill both orders?

Supply network and delivery routesUse the corridor menu to choose a closure; the route table gives exact values.
Available capacityUnreserved serviceCarrying cratesClosed corridor

The search checks every declared single-route closure. It applies the cheapest passing plan only if it fits your budget.

Entire delivery run

Routes show total crates dispatched during this scenario. Scrub the timeline to inspect a particular departure.

THE DISPATCH RECORD

Follow every departure.

One feasible dispatch

The map shows the outcome. Here is one schedule that makes it happen. Select a time to inspect that departure on the board.

The trace accounts for routed stock, not every crate physically on site. Other maximum-delivery schedules may differ.

Operating brief & assumptions

ILLUSTRATIVE SCENARIO

Units, assumptions & what’s left out

Receiving windows and capacity limits are established logistics constraints. This is a smaller network-capacity model, not a vehicle-routing solver. Time-window reference: OR-Tools.

Budget, deadline & network
See why the plan changesThree guided comparisons

Each step loads a built-in example onto the live board. Undo last change steps back one board change, including its network and reservations.

Compare the steps & inspect the evidence

These figures are recomputed by the same finite engine as the board. “Worst case” means the lowest on-time total across ordinary operation and each declared single-corridor closure—not a failure probability. Use the dispatch ledger and “What exactly is being proved?” below the board for exact schedules, bounds and assumptions. These are fictional teaching examples, not operating advice.

What if two corridors close together?
Open the joint-closure workshop ↗

CHANGE THE PROBLEM

Your network, your constraints.

Change the values below, or use JSON to add places, routes, stock, orders and investment packages. Apply makes one reversible change; a rejected edit leaves your current network intact.

Edits invalidate any earlier optimal-plan claim.

FOLLOW EVERY CRATE

The dispatch ledger

Flow / cut certificate
Only positive shipments appear. Unused stock may remain outside the routed plan.

SMALL WORLD, PRECISE PROMISE

Resilient to what, exactly?

One corridor, gone for the run.

The test closes each declared physical corridor in turn, including every directed route in that group. The design is chosen first. Dispatchers then know which corridor is closed and can reroute accordingly.

That is a narrower promise than surviving a surprise failure halfway through a journey. There are no probabilities, repair crews or simultaneous disasters hiding in this model.

The clock is part of the network.

A route takes whole time steps. Crates arriving at time 2 can leave again at time 2. Keeping them until time 3 consumes storage. Capacity applies to each departure, not to the entire day.

Each order is counted once over its arrival window. A crate delivered to one depot is consumed there; it cannot satisfy another order down the road.

Why “least-cost” is earned.

The search checks legal investment subsets in increasing cost order. A design passes only if every order is met in every declared closure case and on the ordinary run. The first passing design is therefore cheapest in this menu, even if it exceeds your budget. Ties are possible.

If the search is cancelled, there is no optimality claim. If you change an input, the old answer is marked out of date.

Not a black box.

Each scenario is an integer maximum-flow problem on a time-expanded network. The ledger conserves stock at every place and time. An independently summed cut with the same value certifies that no more crates can be delivered in that scenario.

Development checks compare the solver against a separately written, individual-journey formulation on bounded cases. They are not external expert validation.

Comparisons have a declared objective.

Compare plans enumerates every legal subset and every declared case. Its efficient set uses two objectives: lower investment cost and higher worst-case total delivery. It does not optimize fairness, travel cost or the probability of a disruption.

Investigation removes one selected package at a time and runs explicitly described input changes. A null result means no change in the displayed measure, not that a relationship never matters.

Scope, exact limits and sources

This fictional instrument accepts up to 12 places, 24 directed routes, 84 stock releases, 24 orders, 6 time steps and 10 binary investments. Quantities are whole numbers, individually at most 1,000,000. Packages may exclude one another. All routes, quantities and costs are declared—there is no inferred geography or real-world calibration.

Stock releases make crates available; the plan need not inject all available stock. Storage limits only the crates routed across a waiting interval. It is not a warehouse audit of all physical surplus.

The method uses standard network-flow ideas, not a new optimization algorithm. Background: Google OR-Tools: network flows and NetworkX flow algorithms. These libraries do not run in this website. No general emergency-readiness or performance guarantee is implied.

Engine . Desktop and narrow-layout checks do not establish physical-phone performance or uncoached visitor comprehension.

KEEP THE EXPERIMENT

A plan you can reopen.

Save the network, selected investments and current closure together. Saved results are a record, not a trusted certificate: loading a file recomputes its deliveries.

Files stay on your device. Changes are held only in this page’s memory. Download anything you want to keep before reloading, navigating away or closing it.

APTIFOLD / AGENTIC AI

What would you like to work through?

Aptifold develops its own agentic AI to investigate questions, test ideas and build working software. This experience is one example of that work.

Explore Aptifold Discuss a question

Company links open in a new tab, so you can keep this experiment open.