01 · Sources
Claims should be checkable.
Essays that depend on current standards, regulations, research, incidents, or technical documentation link to the underlying source where practical. We prefer primary material and distinguish a published rule from a draft, proposal, interpretation, or inference. Links are evidence pointers, not a substitute for accurately describing what a source supports.
02 · Agent assistance
Assistance is disclosed at the level that matters.
Software agents may support research discovery, outlining, comparison, drafting, editing, formatting, link checks, metadata, and publication operations. They do not receive authorship or editorial accountability. Agent output is treated as unverified material until a human editor reviews the argument, sources, language, and publication state.
03 · Review
The editor owns the published claim.
- Factual and time-sensitive claims are checked against available sources.
- Quotations and numerical claims are checked for context and scope.
- Unsupported product, customer, legal, security, and performance claims are removed.
- Internal or restricted company material is not used as public copy.
- Publication surfaces—page, archive, feed, sitemap, and social metadata—are checked for consistency.
04 · Corrections
Correct the record without erasing it.
Material factual errors are corrected promptly. When a correction changes the meaning of an essay, the page should identify what changed and when. Minor typographical, formatting, and broken-link repairs may be made without a note when they do not alter the claim. We do not silently rewrite a published position to make its earlier version disappear.
05 · Responsibility
Human responsibility does not transfer to the tool.
Brian C. Long is the editor of Dispatches. Summit Cognitive is responsible for the publication systems it operates. If an essay is wrong, misleading, improperly sourced, or published in error, “the agent did it” is not an editorial defense.