Provenance Record

Well-structured Analysis piece with clear Overview / What We Know / companion-papers / What We Don't Know / Analysis sections. Technical concepts (Clifford vs. non-Clifford gates, spacetime codes, syndrome post-selection) are explained accessibly without oversimplifying. The closing Analysis section is original synthesis appropriate to the Analysis category and does not introduce unsourced claims.

All 4 sources read in full from local gzipped snapshots (sha256 of each decompressed file matches manifest.json exactly, confirming snapshot integrity). (1) source-0.html.gz - IBM newsroom press release: confirms July 30 announcement, paper title 'Sampling hard circuits with verifiably high fidelity', 70 logical qubits, 2,415 logical two-qubit operations, 468 logical T gates, ~15 minute runtime, and verbatim quotes from Fefferman, Ghosh, and Gambetta. (2) source-1.html.gz - IEEE Spectrum ('IBM Claims Quantum Advantage With New Validation Techniques', Edd Gent, 31 Jul 2026): confirms the 95% confidence framing, the 28% retained-run fidelity figure, the 860x run overhead, the Clifford/non-Clifford gate mechanism, the Qedma and Algorithmiq companion-paper details (35/51/74-qubit thresholds, Quantinuum H2/Helios cross-validation, 56-qubit Algorithmiq circuit), and verbatim quotes from Gambetta, Lindner, Maniscalco, Hangleiter, and Eisert. (3) source-2.html.gz - Phys.org, served via Archive.org fallback per manifest (archive_fallback: true, original returned HTTP 200; snapshot content is the same UChicago-authored press release republished by Phys.org): confirms the same core figures and the exact phrase quoted in the article's Overview ('the point where quantum computers can be confirmed to have outperformed classical computers on trusted computations'). (4) source-3.html.gz - arXiv abstract page for 2607.25941: confirms paper title, all 9 listed authors (Martiel + 8 co-authors including Fefferman, Ghosh, Gambetta), and the fidelity/confidence/suppression figures (0.284 fidelity lower bound, 95% confidence, 10x error suppression) -- but see finding below: the article quotes this abstract with reordered/altered wording inside quotation marks rather than verbatim.

Every number, name, date, and attribution I checked traces cleanly to a cited source: 70 logical qubits, 2,415 two-qubit operations, 468 T gates, ~15 minutes, 0.284 fidelity lower bound, 95% confidence, 10x error suppression, 860x run overhead, 28% retained-run fidelity, the 9-author byline, and the 35/51/74-qubit and 56-qubit companion-paper figures all matched the snapshots exactly. One issue found: a quoted sentence attributed to the arXiv preprint reorders and rewords the actual abstract text (see findings). The underlying facts in that quote are correct; only the verbatim-quote convention is violated, and it is a single, isolated instance in the body (not headline/summary/lead).

Strong Analysis piece with accurate, well-sourced reporting across all four sources including the underlying arXiv preprint. The lone issue -- a paraphrased passage presented as a verbatim quote -- is a single recoverable defect that does not touch the headline, summary, or lead and is honestly addressed via a public corrections record. Approved with corrections.