A *resource estimate* says what a fault-tolerant machine would need to run a program: how many physical qubits, for how long, with how many magic-state factories. It starts from the program's logical counts (its qubits, its T gates, its Toffolis, its rotations at the T count of their synthesised words) and adds the physical model: a code (here the surface code, its distance chosen so that the whole run fails with probability below an error budget) on a physical qubit with a stated error rate and speed.

The engine estimates from the program itself, and every number in the estimate comes from a stated model with its assumptions listed. Every estimate is sealed, and `verify` re-runs it:

```sh runnable
cat > carry.qasm <<'QASM'
OPENQASM 3.0;
include "stdgates.inc";
qubit[3] q;
ccx q[0], q[1], q[2];
QASM
wai-quantum estimate carry.qasm
wai-quantum estimate carry.qasm --json > carry-estimate.json
wai-quantum verify carry-estimate.json
```

The Studio's **Estimate** pane shows the same counts, the estimate on a chosen qubit and the frontier between qubits and time.

Uncomputing work is often necessary (it returns ancillas to |0⟩ so they can be reused), and it costs. This exercise's check runs the estimate on your program, and the result shows the physical figures it gives.

This is the end of the course. The companion course, *Quantum, reconsidered*, takes the same engine into interference-based learning and the limits of quantum advantage.