Further Reading: Quantum Cloud Platforms

Tagged Tier 1 (confident it exists and recommended) and Tier 2 (real and worth seeking, but verify the current version or URL).

This chapter's sources are unusual for the book: the primary literature is vendor documentation, which is accurate, well-maintained, and changes without notice. Read it as a specification, and record the date you read it — which is the same discipline PriceBook.as_of enforces in code.

Platform documentation

• IBM Quantum Platform documentation, particularly the Qiskit Runtime primitives guide (Sampler, Estimator) and the execution modes page (job / batch / session). §39.7's session arithmetic is in that second page, and it is the highest-value operational reading in this chapter. Tier 1.
• AWS Braket developer guide, particularly the pricing page and the device-properties API. Braket is the clearest place to see the per-task + per-shot structure and how it varies by modality. Tier 1.
• Azure Quantum documentation on providers, targets, and the credit model. Tier 2.
• Quantinuum's H-Series documentation on HQC (Hardware Quantum Credits), which prices a job as a formula over qubits, gates, and shots rather than a flat rate. Read it as the third pricing structure — it is the one that most explicitly acknowledges that a shot is not a unit of work. Tier 1.

Primitives and error mitigation as a service

• Qiskit Runtime's Estimator documentation on resilience_level. What mitigation is applied at each level, and what it costs in shots. Chapter 13's techniques, productized — and worth reading precisely because the default is not "none." Tier 1.
• Documentation on measurement-basis grouping in the Estimator. Chapter 24 §24.3's optimization is a platform feature, and whether it is applied is a configuration value you should know. Tier 1.
• Literature on Pauli twirling and probabilistic error cancellation as implemented by providers. Tier 2 — the implementations move faster than the papers.

Benchmarking and hardware metrics

• Chapter 30's reading list applies in full. Quantum Volume, CLOPS, and the metrics vendors publish are the numbers you will be shown when choosing a backend, and Chapter 30 §30.5's refusal to quote a fidelity without naming its statistic is the right posture toward all of them. Tier 1.
• CLOPS (Circuit Layer Operations Per Second) specifically, because it is the one published metric that tries to capture §39.3's throughput question rather than fidelity. Read it asking what it includes and excludes about the queue. Tier 1.
• Vendor roadmaps. Useful for direction, and to be read as marketing about the future rather than data about the present. Tier 2.

Reproducibility

The best material here is not quantum-specific, and that is the point.

• The broader computational-reproducibility literature — pre-registration, artifact evaluation, reproducibility badges in ACM/IEEE venues. The field has already built the machinery §39.8 asks for, and quantum computing has not adopted it. Tier 1.
• Work on machine-learning reproducibility, especially on seed variance and single-run reporting. Case Study 39.2's failure is a known failure mode there with a developed literature. Tier 1.
• Qiskit's backend properties() and target APIs. How to capture the calibration snapshot that §39.8 says you must. Tier 1.
• Chapter 29 §29.3 on stale calibration data, which this chapter's §39.6 is the productized form of. Tier 1.

Cost and capacity

• Papers reporting resource estimates for near-term algorithms — Chapter 36's reading list has the chemistry ones. Read them converting shots to dollars, which almost none of them do, and which §39.5 shows changes the conclusion by 3,718×. Tier 1.
• Anything published on QPU utilization and scheduling. Thin literature, large practical consequence, and §39.3's arithmetic suggests why it matters more than it is studied. Tier 2.

Backward references

• Chapter 13 — the mitigation techniques the Estimator applies for you.
• Chapter 24 §24.3 — the shot budget, and the commuting-group optimization that is now a platform feature.
• Chapter 29 §29.3 — calibration goes stale, which §39.6 measures as layout roulette.
• Chapter 30 §30.5 — never quote a number without naming its statistic.

Forward references

• Chapter 40 — what all of this employs people to do.

Where to go next. If one thing: the execution-modes documentation for whichever platform you use. §39.3's arithmetic says the queue dominates by four to six orders of magnitude, and execution modes are the only lever that addresses it. Nothing else in this chapter's reading will change your wall clock as much.

If two: add your provider's pricing page, read alongside job_cost(). §39.5's 3,718× is not a negotiation outcome — it is arithmetic over a shot count, a circuit duration, and a rate card, and all three are available to you before you run anything.

Then Chapter 40, which closes the book by asking what this field actually employs people to do — and what forty chapters of measurement add up to.