Choose your learning path

Three entry points, by what you already know. Times are honest estimates for a focused session.

Path A — power system researcher, new to quantum computing

You know MATPOWER or pandapower; the word “ansatz” means nothing yet.

Step	What	Time	You leave with
1	10-minute quickstart	10 min	the library runs on your machine; you solved an islanding QUBO
2	Quantum computing for power engineers	10 min	the five terms, the machine model, the honest hardware picture
3	Tutorial 01 — hello, QuGrid	15 min	you can read a Result and explain gap() and success_probability()
4	Tutorial 03 — quantum optimization 101	45 min	you know what QAOA iterates and can interpret its landscape plot
5	examples/03 (islanding) and examples/02 (unit commitment)	20 min	two complete studies you can modify for your own cases
6	Tutorial 05 + examples/08 if you work on ML	45 min	quantum kernels on an N-1 screening task, with the bandwidth lesson

After this path you can formulate your own problem with QUBOBuilder (Tutorial 02 shows every step).

Path B — quantum researcher, new to power systems

You can derive QAOA; you have never seen a bus admittance matrix.

Step	What	Time	You leave with
1	Power systems for quantum researchers	10 min	the vocabulary, the credibility rules, the scale reality table
2	10-minute quickstart	10 min	working install
3	Tutorial 02 — from MATPOWER to QUBO	30 min	how grid constraints become penalties, and what that costs
4	examples/01 (HHL on DC power flow) and Tutorial 04	40 min	the linear-algebra side: conditioning, clock bits, postselection
5	Honest benchmarking	10 min	the reporting standard power venues expect
6	examples/07 (benchmark runner)	15 min	a seed-swept comparison table ready for a paper

After this path you know which grid problems are worth your algorithm ideas — and which comparisons will get a paper rejected.

Path C — you know both; you are here to do research

Step	What	You leave with
1	Problem-to-algorithm cheatsheet	the API map
2	examples/07 + bench module	your experiment harness: tidy DataFrame → LaTeX
3	examples/05 (discretization study)	the encoding-vs-solver error separation, ready to extend
4	Adapters (to_qiskit_operator, to_bqm, to_pennylane)	the same encodings on vendor stacks and hardware
5	Contributing	where a new formulation or solver slots in (one file each)

The library's design contract: formulations own the physics, encodings own the algebra, solvers own the search. If your research adds one of the three, you write one file and the rest of the machinery (references, gaps, benchmarks, plots) applies to it automatically.