Editorial policy

How Gen α AI content is produced

Our research pipeline, where AI is involved, where humans are, and how to get something fixed. Last updated July 4, 2026.

Gen α AI is a research-driven publication about AI engineering, and we use AI systems in producing it. This page describes exactly how, so you can judge the work on its process — not just its polish.

How an article is made

  • Multi-source deep research pass. Every flagship piece starts with an automated deep-research run that fans out across dozens of primary sources — papers, system cards, vendor documentation, benchmark repositories, and engineering write-ups — and returns a cited research brief. Claims without a source do not make it into the brief.
  • AI-assisted drafting. A large language model drafts the article from that research brief under a house style that requires citations, concrete numbers, and explicit uncertainty ("we don't know yet" is an acceptable sentence here).
  • Automated fact-checking editor. A separate editorial pass re-verifies the draft's checkable claims against the live web — links, figures, dates, version numbers, benchmark scores — and flags or removes anything it cannot confirm. Articles that fail this gate are held for review instead of being published.
  • Human editorial oversight. A human operator (Srijan Poudel) sets the editorial direction, reviews held drafts, curates what gets published, and is accountable for everything that appears on the site.

What we don't do

  • We don't publish paid placements, sponsored posts, or affiliate-driven recommendations. There is nothing to disclose because there are no such deals.
  • We don't invent quotes, benchmarks, or sources. If a number appears in an article, it traces back to a source found during research.
  • We don't quietly rewrite history. Substantive fixes update the article's modified date; corrections that change a conclusion are noted in the text.

Corrections

If you find an error — a wrong number, a dead link, a claim that doesn't hold up — email [email protected]. Verified corrections are fixed in the article itself, and the piece's last-modified date is updated. We would rather correct a piece than defend it.

Why we publish this

AI-assisted publishing is common; saying so plainly is not. We think the interesting question is not "was AI involved?" but "was the process rigorous?" — and the only way to let you answer that is to show the process.