Enquire
Menu

Research & insights

Follow the
evidence.

Independent research on AI, cybersecurity, architecture and governance. Explore the methods, trade-offs and evidence behind technology and risk decisions.

Browse the research ↓ Latest edition ↗ Cybersecurity research ↗

Practical research for technology and risk decisions.

Conceptual evidence structure. Sources and analysis support a traceable position.

114 published articles · Architecture / Governance / Method / Data

The lead feature

A closer look at the decision and the evidence behind it.

Featured · Governance · 6 min

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.

19 Aug 2026Read the feature ↗

Governance / Research study

Conceptual study · Trace the evidence.

Recent editions

The latest thinking.

Recent editions, with sources and limitations made explicit.

03 Nov 2026

Governance · 4 min

Share a cyber observation with a route to correct it

Shared cyber observations can speed up local investigations. Preserve their evidence and revision state, and give the receiving organisation its own decision about any response.

Read the edition →
02 Nov 2026

Data · 4 min

Test the data intervention behind a fine-tuning regression

Gradient rankings can help identify training examples worth investigating. Test a named change to the training mix against both the original mixture and a comparable removal before adopting it.

Read the edition →
01 Nov 2026

Architecture · 4 min

Carry the selection from a chart into the next answer

An interactive chart can become the input to a follow-up question. Carry the selected data, grouping and filters so the receiving analysis can answer for the same population.

Read the edition →

The research library

Explore the archive.

Browse by subject or explore earlier editions.

31 Oct 2026 Screen the installed model before comparing quality A quantised model can load and still fail basic tasks. Check the installed file, template and runtime together, then use controlled comparisons to diagnose failures before a wider evaluation. Architecture 30 Oct 2026 Write the event before generating its paperwork A synthetic workplace can test how an assistant reconciles records. Establish the underlying events and expected answers before generating the documents used to score it. Data 29 Oct 2026 Track what changed in a supplier market scan A wider supplier search can find viable offers that already existed. Keep the membership and criteria of each scan so procurement can separate better discovery from real changes in supply. Method 28 Oct 2026 Budget the weight reads in sparse decoding Fixed channel counts can make one part of model decoding easier to budget. Compare sparse and full execution on weight traffic, completed answers and the cost of retaining the fallback. Architecture 27 Oct 2026 Record the analytical choices an agent helps make An agent can influence an analysis without introducing a factual error. Record consequential choices about outcomes, populations and weighting while the work proceeds, so approval can examine the question being answered as well as the calculation. Governance 26 Oct 2026 Choose visual outputs for the job they must do WeVisOne-VL brings points, boxes, masks and transparency mattes into a shared language interface. Choose the required geometry from the image operation, then assess target selection and boundary quality separately before consolidating the workflow. Architecture 25 Oct 2026 Prepare voice lookups with a clear handover Voice agents can prepare record lookups while a caller speaks. Reuse only completed results that fit the final request, and assign unfinished work explicitly so early preparation reduces waiting without mixing incompatible answers. Architecture 24 Oct 2026 Keep forest types visible in parcel summaries New forest maps distinguish land types that a single parcel label can hide. Preserve the class breakdown, smaller patches and unresolved coverage when turning map outputs into a sourcing review. Data 23 Oct 2026 Bind a software proof to the release it supports AI can help construct machine-checked software proofs. Keep each result tied to its binary, entry point and operating assumptions so a later release does not inherit a claim that has not been established for it. Architecture