About & Editorial Standards
What this publication is, who writes it, how sources are handled, and what happens when something here turns out to be wrong.
Who Writes This
Argus is the pen name of the independent writer who researches and writes AGI Doomsday Clock. It is a pen name, not a legal name, and that is stated here deliberately rather than left for someone to discover.
The reason is straightforward: this publication writes critically about large, well-resourced AI companies, and the work is meant to be judged on its sourcing rather than on who is doing it. Every substantive claim on this site links to a primary source so that readers can check the reasoning without needing to trust the author. Pseudonymous writing has a long history in this particular field, and in technical and policy writing generally.
What that pseudonymity does not cover: no credentials, institutional affiliations, or professional history are claimed anywhere on this site, because there are none to claim. This is one person reading primary sources carefully. Where the work depends on expertise the author does not have, it cites people who do.
What This Publication Covers
AI safety and alignment research, dangerous capability evaluations, deception and scheming findings, AI governance and regulation, agentic AI risk, and AGI timelines. Currently 32 long-form briefings, split between evergreen explainers of foundational concepts and analysis of specific incidents and research as they are published.
What it does not cover: consumer AI product launches, funding rounds, or industry personality conflicts, except where there is a genuine safety angle.
Sourcing Standards
- Primary sources first. Where a lab, agency, or research group has published its own account, that document is linked directly rather than a secondary write-up of it.
- Claims are attributed. If a number, date, quotation, or finding appears here, the source it came from is linked in the same paragraph or nearby.
- Verification limits are stated. When a story rests substantially on a single source, particularly a company's account of its own conduct, that limitation is said explicitly in the article rather than smoothed over.
- Quotations are verbatim. Paraphrase is presented as paraphrase, not inside quotation marks.
- Headlines must be supported by the article. A headline here should be provable from the first paragraph. Where a story is being reported elsewhere in more dramatic terms than the evidence supports, that gap is usually the subject of the piece.
Corrections Policy
Errors get fixed, and material corrections are noted in the article rather than made silently. Where a claim in an older piece is overtaken by later evidence, the article is updated and the change is described in the text, so a reader who saw the original is not misled by the revision.
If something here is wrong, send a correction. Corrections are welcome regardless of tone, and a specific pointer to what is wrong and what the correct source says is the fastest route to getting it changed.
Independence
This site takes no funding from AI labs, advocacy organisations, or anyone with a stake in how the subject is covered. There are no sponsored posts, no affiliate links, and no paid placements. If that ever changes, it will be disclosed on this page and in any affected article before the fact.
On the Use of AI Tools
Research and drafting for this site make use of AI tools, which is worth saying plainly on a publication about AI. Every published article is edited by a human, and every factual claim is checked against the primary source it cites before publication. Responsibility for what appears here, including any errors, rests with the author and not with any tool.
Contact
Corrections, tips, and disagreement are all welcome via the contact form. Articles are available by RSS.