Part VIII. Applications and Strategy · Chapter 72
When to Build, Partner, Wait, or Avoid
Strategy is choosing the next action, not choosing a mood. This chapter turns quantum enthusiasm and skepticism into six concrete decisions — build, partner, invest, monitor, wait, avoid — each with its own evidence threshold.
In this chapter 9 sections
Choose the least irreversible action that buys information or capability before the next credible technical gate, using explicit option cost, dependency, deadline, and kill criteria; revisit only when a named observable changes.
Build, partner, wait, and avoid are policies with different reversibility. A small partnership can buy a baseline and integration artifact; a full build consumes organizational attention and creates maintenance obligations; waiting preserves cash but may lose learning time; avoiding closes the option until a named fact changes. The decision record must price those differences.
Decision labels encode different commitments
Each option receives maximum cash, engineering weeks, reusable deliverable, upstream dependency, decision owner, expiry, expansion trigger, and stop condition. An option without a learning deliverable is spending, not experimentation. An option without an expiry is drift. The selected action should be non-dominated under the organization’s stated objective.
- Build
- Own the scarce learning loop and its reusable interface.
- Partner
- Purchase a bounded capability with exit and data rights.
- Wait
- Name the external milestone and monitoring owner.
- Avoid
- Record the failed gate and the evidence that could reverse it.
The option record: cost, learning, dependency, deadline
GAO and the National Academies support a staged posture under current uncertainty. The benchmark and full-stack sources require end-to-end evidence before scale. ACM’s artifact policy makes transferable learning concrete. Official SDK and language references identify the versioned interfaces that a narrow partnership should preserve.
One chemistry proposal under five action policies
A chemistry team can spend four weeks and a capped budget with a partner to produce an active-space choice, classical baseline, and resource sensitivity. A full internal platform build costs six months before resolving those questions. Unbounded waiting produces no record. The bounded partnership is selected because its outputs remain useful even if the hardware thesis weakens.
| Action | Maximum effort | Reusable output | Disposition |
|---|---|---|---|
| Build | 24 engineer-weeks | internal platform | dominated before baseline |
| Partner | 4 engineer-weeks | active-space and baseline dossier | selected |
| Wait | 0 engineer-weeks | none unless monitored | requires expiry |
| Avoid | 0 engineer-weeks | closed option record | only after kill criterion |
Artifact contract. A dated option table with executable dominance and expiry checks. Dominated actions are removed; every retained action has an information gain, maximum spend, owner, deadline, and measurable stop or expansion trigger.
FIELDS = {"owner": str, "cost": dict, "duration": dict, "learning_artifact": str,
"portable": bool, "expiry": str, "stop_trigger": str, "source_id": str}
def choose(dossier, maximum_spend_usd, deadline_weeks):
errors = []
eligible = []
for name, option in dossier.get("options", {}).items():
missing = [field for field, kind in FIELDS.items()
if field not in option or type(option[field]) is not kind or option[field] in ("", None)]
cost, duration = option.get("cost", {}), option.get("duration", {})
if type(cost.get("value")) is not int or cost.get("unit") != "USD":
missing.append("cost_schema")
if type(duration.get("value")) is not int or duration.get("unit") != "weeks":
missing.append("duration_schema")
if missing:
errors.append(name + ":" + ",".join(sorted(set(missing))))
elif (cost["value"] <= maximum_spend_usd and duration["value"] <= deadline_weeks and
option["portable"] and option["learning_artifact"]):
eligible.append(name)
choice = "partner" if "partner" in eligible else (eligible[0] if eligible else "defer")
return {"decision": "invalid" if errors else choice, "errors": errors, "eligible": sorted(eligible)}
dossier = {"scenario": "four-week chemistry information purchase", "options": {
"build": {"owner": "R&D", "cost": {"value": 800000, "unit": "USD"},
"duration": {"value": 24, "unit": "weeks"}, "learning_artifact": "resource ledger",
"portable": True, "expiry": "2026-11-14", "stop_trigger": "resource ceiling missed",
"source_id": "scenario:internal-estimate"},
"partner": {"owner": "science lead", "cost": {"value": 120000, "unit": "USD"},
"duration": {"value": 4, "unit": "weeks"}, "learning_artifact": "portable benchmark",
"portable": True, "expiry": "2026-11-14", "stop_trigger": "no baseline gap",
"source_id": "scenario:partner-SOW"},
"wait": {"owner": "strategy", "cost": {"value": 0, "unit": "USD"},
"duration": {"value": 12, "unit": "weeks"}, "learning_artifact": "scheduled evidence review",
"portable": False, "expiry": "2026-11-14", "stop_trigger": "new evidence release",
"source_id": "scenario:review-calendar"}}}
base_result = choose(dossier, 200000, 8)
bad_options = {**dossier["options"], "partner": {key: value for key, value in dossier["options"]["partner"].items() if key != "owner"}}
bad_result = choose({**dossier, "options": bad_options}, 200000, 8)
tight_result = choose(dossier, 100000, 8)
assert base_result["decision"] == "partner" and base_result["eligible"] == ["partner"]
assert bad_result["decision"] == "invalid" and "owner" in bad_result["errors"][0]
assert tight_result["decision"] == "defer" and tight_result["eligible"] == []
print(f"PASS: 72 option workbook choice={base_result['decision']} invalid={bad_result['errors']} spend100k={tight_result['decision']}")Exact validation command: python3 tools/validate_briefs.py --briefs data/editorial_briefs_64_87.json --from 64 --through 87 --check-rewritten-sources --execute-artifacts
Irreversibility changes the preferred experiment
The partnership contract names files, data rights, environment versions, baseline procedure, and review date. It does not purchase advantage theater or a fixed number of device hours. Expansion requires a specified classical gap and resource regime; failure to reproduce the baseline ends the engagement. This turns vendor access into organizational evidence.
Write a decision that expires cleanly
Prompt. Choose among build, partner, wait, and avoid for a chemistry workflow with uncertain fault-tolerant timing.
Deliverable. Four-option record with cash and engineering cost, learning deliverable, dependency, deadline, kill criterion, and chosen action.
Pass condition. The selected action is non-dominated, has a bounded downside, produces a reusable artifact, and expires on a named date or observable technical gate.
Model answer: partner narrowly, do not scale
Format. Narrow partnership decision capped at an active-space and baseline study.
The model answer selects partner for four weeks, capped at the declared spend, with an active-space and comparator dossier as the deliverable. It rejects build as dominated on information per irreversible cost and rejects unbounded wait because it has no expiry. The action automatically becomes wait if the resource sensitivity misses the target regime, and build is reconsidered only after the evidence gates pass.
Verification. The dominance check rejects a full build and an unbounded wait, while the selected partnership passes cost, learning, expiry, and portability fields.
Companion work
Artifacts for this chapter
These entries resolve to checked-in local source. Commands are reproduced exactly from the chapter manifest, and source-embedded fixtures are exported as direct downloads.
Reproduce or test
python3 tools/validate_briefs.py --briefs data/editorial_briefs_64_87.json --from 64 --through 87 --check-rewritten-sources --execute-artifacts
Provenance
Sources and review
- U.S. Government Accountability Office. Quantum Computing and Communications: Status and Prospects. GAO. 2021government technology assessment
- National Academies of Sciences, Engineering, and Medicine. Quantum Computing: Progress and Prospects. National Academies Press. 2019consensus study report
- Timothy Proctor et al.. Benchmarking quantum computers. Nature Reviews Physics. 2025peer-reviewed perspective
- Lieven M. K. Vandersypen et al.. A look at the full stack. Nature Reviews Physics. 2021peer-reviewed perspective
- Association for Computing Machinery. Artifact Review and Badging. ACM Publications. 2026official reproducibility policy
- IBM Quantum. Qiskit documentation. IBM. 2026official documentation
- Google Quantum AI. Cirq documentation. Google. 2026official documentation
- OpenQASM Technical Steering Committee. OpenQASM 3 specification. Linux Foundation Joint Development Foundation. 2026official technical specification
The load-bearing claims in the chapter are mapped inline to this registered source set. A citation supports only the bounded claim beside it.