Research & insights
Follow the
evidence.
Independent thinking on AI architecture, governance and risk. Explore the methods, trade-offs and evidence behind better decisions.
Practical research for better AI decisions.
The lead feature
In focus.
A closer look at the decision and the evidence behind it.
What the EU AI Act actually asks of a retrieval system
A retrieval system needs recoverable evidence when its answers inform consequential decisions. Establish the applicable obligations, then preserve the sources, system versions and human actions needed to assess what happened.
Conceptual study · Trace the evidence.
Recent editions
The latest thinking.
Recent editions, with sources and limitations made explicit.
Give long-running agents a working account of the task
New research on explicit belief states gives agents a way to track what remains unknown and recognise stalled progress. The opportunity is a more useful investigation workflow, with a measurable cost for maintaining that state.
Hold out the customer before funding the rollout
Fresh tickets from familiar accounts do not establish performance on unseen customers. Match the evaluation boundary to the proposed rollout and report the evidence for each population separately.
Restore the AI service from a compatible record
Approve an AI recovery plan against a restored service that completes useful work. Test document and index compatibility, reconcile current access permissions and state the recovery scope that the evidence supports.
The research library
Explore the archive.
Browse by subject or explore earlier editions.
Keep reading
Keep the evidence close.
New research, useful methods and regulatory changes worth your attention. Follow the feed, or ask for the research notes by email.
RSS needs no email address. Email subscriptions are handled directly, without a third-party form.