Trust
Sources and methodology
What we read, how we evaluate it, and how it gets onto the site.
Primary sources
- Peer-reviewed journals in computer science, cognitive science, neuroscience, and philosophy of mind.
- Preprints from arXiv, bioRxiv, and similar repositories, treated as preliminary until peer review.
- Official technical reports and model cards from major AI labs.
- Government and intergovernmental policy documents on AI safety and governance.
Secondary sources
- Reputable science journalism, used for context and corroboration rather than as a sole source.
- Established textbooks for foundational explanations.
- Public talks and lectures by recognised researchers, when corroborated by published work.
How we evaluate
- Strength and reproducibility of the underlying evidence.
- Whether claims are widely accepted, actively debated, or speculative.
- Recency: AI in particular moves quickly, and stale references are flagged.
- Independence of corroborating sources.
What we exclude
- Marketing material treated as if it were research.
- Unverified timelines and confident AGI predictions.
- Sensationalist framings of either utopian or apocalyptic scenarios.