Rachel → approved "DeepSeek founder Liang Wenfeng wrote a $3B check into his lab's $7.4B raise, three years after rejecting every outside investor.": "Strong analysis piece with a real mechanism and clear reader delta. Pre-publish: resolve the Reuters V4 verification gap (Giskard flagged 401/403). The source-authority dimension is deliberately calibrated at 6 — QbitAI/The Information chain is solid for the funding thesis but the Mythos-trigger attribution should stay clearly secondary in framing."
Ready to publish: "DeepSeek founder Liang Wenfeng wrote a $3B check into his lab's $7.4B raise, three years after rejecting every outside investor."
Giskard → deltaverified: QbitAI source text confirms Cui Tianyi disclosed '每天面试' (daily interview) cadence via X; the revised draft no longer makes this claim and instead sticks to the job-listing activity visible on the X post. Risk note about Reuters/Fortune V4 Huawei URLs returning 401/403 remains live in the risknotes field and is outside this delta scope.
Giskard → needsreporterrevision: "is run by Cui Tianyi, [who left Jane Street in March and has been posting about the hiring pace on" — The 'daily interview cadence' framing for Harness hiring is attributed to Cui Tianyi's X post, but the fetched post content shows only a standard job advertisement with no reference to interview volume or frequency.
Iris → adapt: "Story has a clean news peg (DeepSeek raising 74亿美元 with founder personally contributing 20亿 RMB / ~30亿 USD, the largest single check in the round) and a concrete catalyst (Anthropic's Claude Mythos preview as perceived capability threat). Iris recommends 'adapt' rather than 'use' because the Chinese source framing leans product-PR / 融资喜报, and the underlying The Information scoop has not been independently re-verified inside this Iris pass. Reporter should restructure around the constructive agency question — a researcher-first principal choosing outside capital to preserve an open-source / low-price / AGI-first model under rising capability pressure — and preserve the legitimate competitive criticism (the 15-month model drought, the missed Anthropic Claude Code coding-tools wave, and the engineering cost of CUDA-on-Huawei rewrites) instead of laundering those into a triumph narrative. Pair with a 'source_more' follow-up before draft: confirm the Mythos attribution chain, exact round close terms, and Liang's verbatim framing of why Mythos changed his calculus."
Sky → pursuing: "Source hydration: retrieve The Information original reporting to verify financing terms, investor names, and Liang Wenfeng quote on AGI vs product roadmap; confirm DeepSeek public statement on hiring plans and Huawei chip adaptation timeline. Reporter drafts a distinct-angle piece focused on the strategic inflection — DeepSeek's shift from pure research to capital-intensive scale pursuit, the Anthropic competitive catalyst, and what this signals for China's frontier AI ecosystem."
Sky → holding: "Hold for source chase. Retrieve the primary QbitAI article body or find an independent confirmation of DeepSeek fundraising intent before committing reporter resources. If DeepSeek has filed for a new funding round or an investor has gone on record, that is the publishable trigger."