What the literature establishes, and where it stops
A close reading of the papers behind the Stargazer v2 argument: what each one shows, read from the PDF with the page of every number, and why together they leave the problem open.
- Thesis
- Four groups
- Synthesis matrix
- Papers
- The gap
- What the reading changed
- Open risks
- Reading order
- MSR literature
The thesis the reading has to support
Four groups, four objections
The library is not organised by topic for convenience: each group answers a different objection a reviewer will raise.
Synthesis matrix
The six measurement and estimation papers, the project's own pilot bench, and what a release gate would need. Read across the last row: no paper predicts every cost dimension across a heterogeneous fleet, at release time.
Each cell carries the page (or slide) of the PDF it was read on; the pilot row links to the Experiments page.
The papers
One card per paper, in reading-library order. Small grey marks after a number give the page of the PDF it was read on. Links go to the DOI, arXiv or the paper's official page.
What no paper covers
The open risk to this statement is listed under open risks.
What the close reading changed
Reading the PDFs, rather than abstracts and summaries, corrected several claims already recorded in this repository. The files listed are not edited by this page; follow-up changes are tracked separately.
Open risks and reading caveats
Suggested reading order
Relation to the MSR-study literature
The mining study rests on a second body of work, on bugs in deep-learning and LLM software. Those papers are not part of this close reading; their roles below are summarised from the bibliography. Group D above is where the two bodies of work meet.
Full bibliography, by theme: References. How the literature backs the search terms: Literature Validation. The Round 1 summary of this library: Dossier, §7.