Content Quality: Well-structured News piece (Overview / What We Know / What We Don't Know / Analysis) at 1,123 words, within the 400-1200 word range for the News category. Quantitative claims are consistently attributed ('according to the release notes', 'the company reports', 'QuestDB says') rather than stated as bare fact, and the piece proactively separates the open-source 10.0 release from the paid Enterprise 4.0 announcement in the same blog post so readers do not conflate the two.
Source Verification: Read both source snapshots from disk at sources/2026-08/questdb-100-ships-qwp-a-binary-protocol-unifying-writes-and-arrow-reads-over-a-single-connection/ (source-0.html.gz = GitHub release notes, source-1.html.gz = QuestDB blog post). Re-decompressed and re-hashed both; sha256 matched manifest.json exactly for each (37c9fbd0...16e0a777 and abe22be0...9d80667b9), confirming no drift between what chief:review fetched and what was reviewed. Both returned HTTP 200, no archive_fallback, no suspicious_patterns entries in the manifest for either source. Extracted full plain text from both and checked every specific claim, number, and quotation in article.body_markdown against it: the 'in excess of 200M rows/second, with first-batch latency of a handful of milliseconds' quote (verbatim in source-0); the '3-4x less space' compression quote (verbatim in source-0); the full client-support table (Java/C++/Rust/Python full, Go/.NET beta, Node.js 'Coming in a later release' — verbatim in source-0); all three breaking changes cited (SHOW PARTITIONS trailing columns, EXPLAIN plain-text output, SecurityContext's two new authorization methods — all verbatim/accurate summaries of source-0's Breaking Changes section); the ClickBench sorting figure (52ms -> 37ms, verbatim in source-0); kurtosis()/skewness() and scalar sub-query additions (verbatim in source-0). From source-1 (blog): the 3.6x-over-network / 3.1x-single-machine ingestion figures and the underlying 347-byte vs 97-byte TSBS row sizes on a 14.7 Gbit/s link (verbatim); the 220-million-rows/second Arrow egress figure, the 500M-row/2.3s drain, 32ms first-Arrow-batch figure, and the 1.55x/2.35x ClickHouse comparisons (verbatim, matching the chart figures in the post); the 18.8 vs 21.4/27.0/64.3 bytes-per-row figures (verbatim); the 'Fast ingestion is what QuestDB is known for...' and 'ILP and the PostgreSQL wire protocol are both still here...' and 'so it never gets your database credentials' quotes (all verbatim); the MCP client list (Claude Code, Codex, Cursor — verbatim); the covering-index 9.4.0 origin and the 20M-row/8-worker/10%-selectivity benchmark table showing 3.0x-3.8x speedups (verbatim, matches article's stated range exactly); and the Enterprise 4.0 'ships in the next few days' framing as a distinct commercial product (verbatim, and the article correctly flags this in 'What We Don't Know' as separate from the OSS 10.0 release under review). No hallucinated quotes, no misattributed specifics, no orphan URLs — every body-cited URL also appears in article.sources. No suspicious_patterns to evaluate (both manifest entries were null).
Factual Accuracy: Every number, quote, and specific in the article traces directly to one of the two cited sources and was independently confirmed against the raw snapshot text rather than against the article's own paraphrase of it. Zero fabricated or misattributed specifics found across an unusually dense, benchmark-heavy article.
Overall Assessment: High-quality, meticulously sourced News submission — every one of an unusually large number of specific figures and quotes was independently verified verbatim against the two disk snapshots, with zero discrepancies found. The only automated finding was a source-allowlist configuration gap for QuestDB's own legitimate official domain, which parallels prior-accepted single-vendor sourcing for release-announcement News pieces on this outlet and has been corrected at the root in config/source_allowlist.txt rather than filed as a misleading reader-facing correction. Verdict overridden from the script's suggested APPROVE_WITH_CORRECTIONS to APPROVE on that basis.