1. Fully autonomous publication
Briefings on McGauley Labs are drafted and published end-to-end by a multi-agent software pipeline running on a cron schedule. There is no per-briefing human approval. Brian McGauley does not read every draft before it goes live.
The accountability model is the editorial guardrails, not per-piece review:
- The agent prompts and the public style guide govern voice, sourcing, attribution, what is and isn’t reportable, and the disclosure language on every briefing.
- Brian owns those guardrails. Iterating on them is how the publication’s standards are enforced.
- When something publishes wrong, the failing briefing is corrected and the underlying prompts or style guide are updated so that failure mode stops recurring (see §7).
2. Sourcing
Every briefing is built from publicly available primary sources — official announcements, research papers (arXiv, conference proceedings, lab blogs), regulatory filings, SEC documents, court records, company press releases, and reporting from established news outlets. Each briefing publishes the full list of source URLs at the bottom of the article in a sources block.
The style guide prohibits knowingly drawing on leaked, hacked, or non-consensually-shared material, and prohibits citing anonymous sources we cannot verify ourselves.
3. Verification
For factual claims, the style guide requires either a primary source (the company or person being described) or at minimum two independent secondary sources. Numerical claims — funding amounts, benchmark scores, model parameter counts — are cross-checked against the original announcement by the AnalyzerAgent before they reach the drafting step.
4. The drafting pipeline
Each briefing is produced by a sequence of named software agents:
- SearchAgent / ResearchAgent — query news APIs, RSS feeds, and research repositories for the latest material.
- CollectorAgent — fetch and normalize full article content.
- DBUpdaterAgent — deduplicate against what we have already covered.
- AnalyzerAgent — extract significance, priority, and category signals.
- CategoryAssignmentAgent / GroupingAgent — group related stories into a coherent edition.
- BlogAgent / WeeklyWriterAgent — produce the draft briefing.
- ArtistAgent — generate the featured image.
- FormatterAgent — finalize layout and metadata.
- EmailAgent — dispatch the email edition.
- Orchestrator — coordinate the pipeline and publish.
The drafting agents use Google Gemini as the underlying language model. The agents pull the current style guide at the start of every run and inject the relevant sections directly into their prompts. See /ai-disclosure for the agent-by-agent breakdown.
5. Attribution
Every cited source is linked. Quotation marks indicate verbatim quotes; everything else is paraphrase or analysis. The style guide forbids fabricated quotes, fabricated sources, and the use of AI-generated likenesses of real people in featured images.
6. Independence, conflicts, and advertising
McGauley Labs is independent and unaffiliated with any AI lab, vendor, fund, or institution. We do not accept sponsored briefings, paid coverage, or affiliate placements. If Brian McGauley holds a financial position in a company being covered, that position is disclosed inline in the briefing.
The site is funded by display advertising (Google AdSense) and reader subscriptions. Ad placement is programmatic; advertisers have no editorial influence and cannot affect which stories the pipeline covers.
7. Corrections
When a briefing is wrong, two things happen:
- The published briefing is corrected, with an editor’s note recording what changed and when.
- The agent prompts and/or the style guide are updated to prevent that class of error from recurring. This is the real corrections mechanism for an autonomous pipeline.
See the corrections policy for how to report something and what timeline to expect.
8. Updates to this policy
Material changes to this policy are summarized at the bottom of the page when they happen. This policy was last reviewed on 2026-05-28.
Contact
Editorial questions, corrections, security: brian@imagi-narii.com.
Support and account help: support@imagi-narii.com.