Sync is an editorial brief built on a simple discipline: every claim traces to a named public document. Earnings transcripts, regulatory filings, official press, conference talks, the trade publications that cover this beat seriously. If we can't cite it, we don't write it.
Three pillars.
Source discipline
Every story lists its sources at the bottom. Every link goes to a primary document — SEC EDGAR, an investor-relations page, a conference transcript, the regulator's own publication. We don't link to summaries of summaries.
Editorial fact-check
Every draft passes through a quoted-span verification step before it ships. If a quote can't be located verbatim in a cited source, the quote comes out and the source is reviewed. About 38% of first drafts don't clear this step. The discipline is the value.
Correction posture
When we get something wrong, we say so and fix it. The correction stays in place; the original isn't memory-holed. See the corrections page for the running record.
How a story gets made.
- Source intake. 117 sources are tracked at a 30-second cadence — earnings transcripts, 8-K and 10-K filings, official press releases, conference talks, the regulators' own publications, and the trade press that covers this beat seriously. Items are deduplicated against URL hashes.
- Editorial review. A first read for relevance — does this affect the C-suite technology beat we cover? — and for citation density. Items without a primary-document anchor are dropped.
- Drafting. The story is written against the source set, with every claim mapped to a citable URL. The draft includes the candidate sources list as inline references.
- Quoted-span verification. Every direct quote in the draft is checked against the originating document. Quotes that can't be located verbatim are removed and the underlying claim re-examined. About a third of first drafts fail this step on the first pass.
- Editorial fact-check. A second read for narrative coherence, for unsupported inferences, and for the difference between what the documents say and what we are tempted to say they say. Anything that can't survive that read is rewritten or cut.
- Publication. The story ships with the source list visible, the date stamped, and the section accent applied. The Brief envelope adds context and forwards to subscribers.
The AI Ratings rubric.
The weekly leaderboard on /pages/implement.html is computed from public benchmarks and provider-published model cards. The composite score is a 60/40 weighted blend of capability and cost-of-capability:
- Reasoning — MMLU 5-shot, drawn from each provider's own model card or third-party HELM reporting.
- Code — HumanEval pass@1 plus the latest SWE-bench Verified score where available.
- Math — GSM8K and MATH, weighted equally.
- Long context — the longest published recall benchmark (RULER, LongBench, or the provider's own).
- Instruction-following — IFEval and MT-Bench-2.
- Multilingual — MGSM averaged across the published language set.
The composite score updates every Monday at 06:12 UTC after the weekend's published-benchmark batch is reviewed. Models are added to the leaderboard when they reach general availability; they are removed when their provider deprecates them. We never accept payment for inclusion, exclusion, or ranking position.
Sourcing policy.
Eligible sources, in order of preference:
- Primary documents — SEC filings (10-K, 10-Q, 8-K, DEF 14A, S-1), earnings-call transcripts, regulator publications (FedRAMP marketplace, EU Commission, FCC, FTC, OCR, CISA), corporate press releases on the company's own newsroom, conference talks on the conference's own publication channel.
- Trade publications — named, sourced reporting from the trade publications that cover this beat seriously (CIO.com, CIO Dive, The Information, WSJ, Bloomberg, Reuters, FT, Axios, The Register, ComputerWeekly, SemiAnalysis, Stratechery).
- Public testimony and academic publications — congressional testimony, peer-reviewed papers, working papers from named institutions.
Ineligible: anonymous quotes the original publication does not name; scraped private content; LinkedIn material that is not publicly posted; paywalled content that we cannot ourselves access.
What we won't do.
- No paid placements. We don't run sponsored stories, sponsored newsletters, or "branded content." There is no path for a vendor to pay to appear in any Sync surface.
- No anonymous sourcing. We do not publish anonymous quotes. Every quoted span traces to a named public document.
- No tracking pixels in the email. The daily Brief carries no open-rate beacons or click-tracking pixels. We cannot tell whether you opened it.
- No paywall. Every story on the site is free to read. The Brief is free to subscribe to.
- No share-to-read. No quote-to-unlock. No email-to-unblock. The work is the work.
Corrections, removals, and the right to be wrong in public.
When we publish something incorrect, we correct it on the page where it appeared, with a dated note explaining what changed. We don't memory-hole. The original framing stays visible alongside the correction so the reader can see the full record.
For corrections to stories: see the corrections page. For data-subject access requests: see the DSAR portal.
Contact
Editorial: desk@sync.community. Corrections: corrections@sync.community. Privacy and DSAR: privacy@sync.community. Everything else: hello@sync.community. A real person reads each address.