Editorial Standards

How Omniscient is reported and edited

These are the rules the publication holds itself to. They are public because credibility on a beat like this depends on being explicit about how the work is done, not on asking readers to take it on faith.

Sourcing

Primary sources first. For every fact-bearing claim, Omniscient cites the primary source where one exists publicly: the paper, the filing, the company post, the transcript, the dataset. Wire and secondary reporting are used in two cases: when the primary is paywalled and the wire is the most-accessible report of record, or when the wire itself is the original (a Reuters or Bloomberg scoop, for example). The default is to read the original document and link readers to it.

AI assistance

Omniscient Media is one editor using custom, self-built AI pipelines to do work that previously required a newsroom. The editor - Noah Ogbi - picks the stories, sets the thesis, chooses the primary sources, structures the argument, and writes the final voice. Custom pipelines built around frontier models do the surrounding labor: scanning for stories, pulling primary sources, aggregating research, producing first-draft outlines and summaries, and stress-testing arguments before publication. Every piece is read, edited, and signed by the editor before it ships. The analytical thesis is editorial judgment, not model output. We disclose this for the same reason we cite primary sources: the credibility of the publication depends on being explicit about what it is. There isn’t really a transparent AI-assisted solo publication on this subject, and we’d rather claim that position than try to fit in the box of a traditional publication. Omniscient Media is a publication of ForeverBuilt, LLC.

Concretely: pipelines scan sources, pull and organize primary documents, aggregate research, produce first-draft outlines, and stress-test arguments. Pipelines do not write the final published voice, choose the editorial thesis, or decide what counts as a primary source. Those are author decisions, and they are the part of the work that determines whether a piece is worth publishing.

Editorial judgment

Author judgment is the load-bearing element of the publication. The pipelines do research; judgment is what makes a piece publishable. The clearest place that judgment shows up is the closer. Every Broadsheet piece is held to a four-question rubric before it ships:

Does the closer name a specific stake, not a generic implication?
Does it return to the lede’s claim and either reinforce or refine it?
Is there a quotable sentence the piece could be remembered by?
Does it tell the reader what to watch for next?

A piece that cannot answer yes to all four gets rewritten or held.

Corrections

Errors are corrected promptly and visibly. Material corrections are noted in the affected post, dated, and added to the public corrections log. The full policy, including how to report an error and what counts as a correction versus a clarification, lives at Corrections.

Pre-publish checklist

Every piece passes through an eight-item checklist before publish. The checklist is enforced inside the editor: four items run as automatic checks against the post’s metadata, four are confirmation prompts the editor ticks. Items unresolved at publish-time trigger a warning; the editor can override, but every override is a conscious call.

Primary-source verification. Every fact-bearing claim links to or cites a primary source unless covered by the exceptions in the sourcing policy above. Run via the Final Draft Check action in the Chief-of-Staff workspace before confirming.
Closer rubric. The closer passes the four-question self-edit above: specific stake, returns to the lede, quotable sentence, what to watch for next.
Read-time populated. The post displays a read-time estimate. Computed automatically at publish time using the 230 wpm formula; manually set values are preserved.
Byline correctness. The author byline reads “Noah Ogbi” exactly, with no defaults, abbreviations, or placeholder values.
AI-assistance disclosure. The post uses standard AI assistance as described above. If any piece uses AI assistance beyond that policy, it carries an inline flag noting where and how.
Hero image attribution. If the cover is AI-generated, the model name is recorded so the public attribution affordance renders correctly. Posts without a cover image pass automatically.
Corrections cross-check. The post does not invalidate any prior claim on the site, or the prior claim has already been corrected in the corrections log.
Tags selected. At least one canonical tag applied. Tags are limited to the subject taxonomy and the eight-company allowlist (enforced by the tag table; the editor can only pick from those).

After publishing

One post-publish check, run as a smoke test outside the editor:

Initial-HTML smoke check. After publish, the live URL is fetched and a distinctive sentence from the body is confirmed present in the initial HTML response. Bodies must be in the page itself, not injected by client-side JavaScript. Catches the rare rendering-pipeline regression where the published page ships an empty article shell.

Conflicts and disclosures

Personal conflicts, including equity positions, prior employment, and active product work, are disclosed on Noah’s author bio. Publication funding: Omniscient Media is a publication of ForeverBuilt, LLC. The publication has no advertisers, sponsors, or paid placements.

Updates to this page

Last updated: 2026-05-25. Material changes to these standards are logged here.

Omniscient Media is a publication of ForeverBuilt, LLC.

Sourcing

AI assistance

Omniscient Media is one editor using custom, self-built AI pipelines to do work that previously required a newsroom. The editor - Noah Ogbi - picks the stories, sets the thesis, chooses the primary sources, structures the argument, and writes the final voice. Custom pipelines built around frontier models do the surrounding labor: scanning for stories, pulling primary sources, aggregating research, producing first-draft outlines and summaries, and stress-testing arguments before publication. Every piece is read, edited, and signed by the editor before it ships. The analytical thesis is editorial judgment, not model output. We disclose this for the same reason we cite primary sources: the credibility of the publication depends on being explicit about what it is. There isn’t really a transparent AI-assisted solo publication on this subject, and we’d rather claim that position than try to fit in the box of a traditional publication. Omniscient Media is a publication of ForeverBuilt, LLC.

Editorial judgment

Does the closer name a specific stake, not a generic implication?

Does it return to the lede’s claim and either reinforce or refine it?

Is there a quotable sentence the piece could be remembered by?

Does it tell the reader what to watch for next?

A piece that cannot answer yes to all four gets rewritten or held.

Pre-publish checklist

Primary-source verification. Every fact-bearing claim links to or cites a primary source unless covered by the exceptions in the sourcing policy above. Run via the Final Draft Check action in the Chief-of-Staff workspace before confirming.

Closer rubric. The closer passes the four-question self-edit above: specific stake, returns to the lede, quotable sentence, what to watch for next.

Read-time populated. The post displays a read-time estimate. Computed automatically at publish time using the 230 wpm formula; manually set values are preserved.

Byline correctness. The author byline reads “Noah Ogbi” exactly, with no defaults, abbreviations, or placeholder values.

AI-assistance disclosure. The post uses standard AI assistance as described above. If any piece uses AI assistance beyond that policy, it carries an inline flag noting where and how.

Hero image attribution. If the cover is AI-generated, the model name is recorded so the public attribution affordance renders correctly. Posts without a cover image pass automatically.

Corrections cross-check. The post does not invalidate any prior claim on the site, or the prior claim has already been corrected in the corrections log.

Tags selected. At least one canonical tag applied. Tags are limited to the subject taxonomy and the eight-company allowlist (enforced by the tag table; the editor can only pick from those).

After publishing

One post-publish check, run as a smoke test outside the editor:

Initial-HTML smoke check. After publish, the live URL is fetched and a distinctive sentence from the body is confirmed present in the initial HTML response. Bodies must be in the page itself, not injected by client-side JavaScript. Catches the rare rendering-pipeline regression where the published page ships an empty article shell.