Research / Methodology
How this research was conducted
Four independent, parallel research passes, each required to open primary sources directly rather than trust search-result snippets.
The primary-source rule
No claim in this research is based on a search-result snippet alone. Every entry in the claims table and evidence record traces to a source that was actually opened and read: an official GitHub repository (README, CI workflow YAML, release history), an arXiv paper (abstract and, where needed, full text), or an independently-resolved DOI record (e.g. a Zenodo software deposit). Where a source could not be opened or a claim could not be confirmed, it is marked UNVERIFIED or NOT FOUND AFTER SEARCH — explicitly distinguished from a claim that was actively checked and found false.
Four parallel research threads
The twelve original claims were grouped into four independent verification passes:
- MQT QCEC and MQT Debugger — official repositories, CI configuration, and associated papers.
- The 31% practitioner statistic and the two originally-alleged arXiv papers.
- Q-Trace / other quantum debugging systems, bug/fault corpora, and state-aware or autonomous testing tooling.
- GitHub and PyPI search for pytest plugins, GitHub Actions products, and cross-SDK/version regression tooling.
Each thread logged every search query it ran and what it returned — the full log is preserved in research_log.md in the GitHub repository.
Status categories
Every claim was assigned one of five statuses:
- Confirmed — verified directly against a primary source with no material ambiguity.
- Partially confirmed — a primary source was verified, but either the original claim referenced something not independently confirmable (e.g. unsupplied arXiv IDs), or the finding only partially supports/contradicts the original framing.
- Unverified — a claim that could not be checked against any primary source.
- False / contradicted — actively checked and found not to hold as originally stated.
- Not found after search — a reasonably broad search turned up nothing, but this is explicitly treated as absence of evidence, not proof of non-existence.
Confidence ratings
High confidence means a primary source was opened and directly inspected (repo code/CI YAML, full paper text/abstract, or an independently-resolved DOI record) with no material ambiguity in interpretation. Medium confidence means a primary source was opened, but either the claim required reconstructing candidates without a pre-given reference to confirm or deny against, or the finding is an absence result that a differently-worded search might overturn.
Known limitations
This was an English-language search across arXiv, GitHub, and PyPI. It did not include a systematic non-English literature search, direct outreach to authors, or an empirical prototype comparison. These and other open items are listed in full in Section 14 of the gap analysis.
Full research log and raw notes: sadeqisaidmohaddes-star/quantum-cicd-research — research_log.md, raw_notes.md