Editorial Standards

What this page is for

Oxford Ledge publishes essays, lesson modules, and data analysis under one editorial standard. This page exists so a reader can audit how a given piece was written and edited before deciding whether to trust it. The standards below are the same ones a contributor signs onto before a single word ships.

If a piece appears on Oxford Ledge that violates a standard here, it’s a mistake, not a policy change. Email editorial@oxfordledge.com and we’ll either correct it or retract it with a changelog entry.

1. Named authors, edited prose

Every long-form essay on Oxford Ledger (our weekly editorial series) is written by a named person and reviewed by at least one other named person before publication. The author byline appears at the top of the piece; the editor of record appears in the footer.

Lesson modules in the LEARN library are authored by the editorial team and revised after pedagogical review. Modules that touch dated material (tax thresholds, statutory limits, regulatory dollar amounts) carry an internal REFRESH_BY marker and surface a Last reviewed footer to readers.

2. No AI-generated prose

We do not publish AI-generated essays, lessons, or analysis. We do use AI to help with research, outline-generation, and code — the same way we use a search engine or a spreadsheet. What ships is written by a human and rewritten by an editor.

The reason is straightforward: AI prose, even when factually accurate, tends to flatten the texture that a careful reader uses to detect when an author is uncertain, hedging, or surprised. We’d rather publish less and have it carry a real point of view.

The product’s AI features (Ask AI, quiz feedback, RAG-grounded answers) are clearly labeled as such, and the underlying model is named in the chat header. Those are tools you operate. The editorial side is a publication.

3. Sourced citations on numerical claims

Any dated or numerical claim in a long-form piece carries a citation in the form of a hyperlink to the primary source plus an access date. When the primary source is paywalled or behind a login, we link to the most stable secondary source we can find and note the gap.

Examples of primary sources we cite by default:

• SEC EDGAR for company filings (10-K, 10-Q, 8-K, Form 4, 13F)
• Federal Reserve Economic Data (FRED) for macro series
• U.S. Treasury Direct for yield curves
• TRACE-reported bond trade data for corporate-bond transactions
• IRS, SBA, and other federal agencies for statutory thresholds

When we reproduce a chart from another publication, we link to the original and credit the author.

4. REFRESH_BY as a policy, not just a CI check

Time-sensitive content (annual IRS thresholds, periodic statutory caps, regulatory dollar limits) carries a REFRESH_BY date in the source code that fails our continuous-integration pipeline if it’s past due. This page itself carries one (next review: 2026-08-13).

The CI check catches stale data. The editorial policy is broader: when we re-review a piece, we update both the prose and the source links if they’ve moved. A re-review never silently bumps the date forward without a corresponding read of the underlying source.

5. Disclosure: what data is licensed vs. public

Most of the data on Oxford Ledge comes from primary public sources we re-host: SEC EDGAR filings, FRED macro series, Treasury Direct yield curves, TRACE-reported bond trade data. We pay for one commercial data feed (Financial Modeling Prep, for company fundamentals and analyst estimates), and we’re evaluating one more (Tradier, for options chains).

The Trust Dossier page lists which dataset each surface relies on, the freshness target, and recent drift catches. The SLA page gives uptime + freshness targets per dataset.

6. Corrections and changelog

When we change a published essay, we leave the original sentence in place and append a dated correction note below it. When we change a methodology, we publish the change in the methodology page’s changelog.

The current published methodology pages:

• Value Creation Score — the 5-component composite (Research Breadth / Depth / Education / Consistency / Community), the soft-log normalization, the 90-day window, the named levels.
• Mastery thresholds — the 5-layer Khan-canonical ladder (Attempted / Familiar / Proficient / Mastered / Enduring), the 70% remediation trigger, the EMA score + calibration + decay rules.
• Insider activity chip — the 90-day Form 4 window, the 1%-of-float threshold, the buy/sell transaction-code mapping, the em-dash fail-closed behavior.

We don’t silently rewrite history. If a chart was wrong, the new chart appears next to a one-line note that says what changed and when.

7. Editorial brand

Oxford Ledger is the canonical name of the weekly long-form series. The lesson library, the data pages, the AI features — those are product surfaces. Oxford Ledger is the publication that lives on top of them.

The Ledger is named for a small writing desk — a ledge on which a notebook rests — and for the double-entry tradition that the platform tries to honor. Every figure is recorded; every claim has a counter-claim that the reader can examine.

Contact

Editorial: editorial@oxfordledge.com · Corrections: same address with subject line beginning Correction · Data complaints: /data-integrity

This page is subject to and should be read alongside our Terms of Service and Privacy Policy. The non-editorial product disclaimers live at /disclaimer.