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.

104 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.

24 Oct 2026

Data · 6 min

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.

Read the edition →
23 Oct 2026

Architecture · 6 min

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.

Read the edition →
22 Oct 2026

Governance · 6 min

Assess what happens to customers excluded from AI

An AI safeguard can reduce problematic answers while leaving redirected customers without useful support. Assess the alternative route and retain excluded customers in the service outcome record before expanding automation.

Read the edition →

The research library

Explore the archive.

Browse by subject or explore earlier editions.

21 Oct 2026 Retrieve the correction with the claim Enterprise search can retrieve an accurate quotation from a superseded document. Keep correction evidence attached to the claim, and establish whether the question asks for the current position or an earlier recorded state. Data 20 Oct 2026 Tune speculative decoding for the load it will serve A diffusion drafter can accelerate a language model by proposing tokens in parallel. Choose its draft length against useful output and request latency at the intended load, because accepting more tokens per round can still make the service slower. Architecture 19 Oct 2026 Share a desktop through explicit targets and handovers OpenGCUA explores agents that work across visual and programmatic interfaces. Sharing a working desktop requires explicit application targets, bounded handovers for foreground input and a separate way to handle competing edits. Architecture 18 Oct 2026 Keep extracted mentions separate from catalogue matches CalOn offers compact, schema-guided extraction with direct links to source text. Use that capability to build a mention layer, then resolve catalogue identities separately so a correct extraction does not become an unsupported business record. Data 17 Oct 2026 Preserve forecast paths when planning for cumulative demand PathFM generates complete future trajectories for forecasting. To use them for cumulative demand or peak-load decisions, preserve each path through the service interface and test the quantity the operational decision actually consumes. Architecture 16 Oct 2026 Give video memory a route back to the footage DeltaStreamer links video summaries to historical frames so later questions can recover omitted details. A useful deployment needs an explicit agreement about retained footage, permitted replay and how long an answer may take. Architecture 15 Oct 2026 Adapt a language model through a versioned corpus table COUNT introduces a way to adapt a language model by changing corpus-count tables without retraining its weights. Treat the table, tokenizer and mixing settings as a release unit, with their own evidence and rollback route. Data 14 Oct 2026 Hold the model constant when evaluating geospatial agents A new Google study separates specialist geospatial tools from the orchestration layer around them. Buyers should compare both against expert-defined workflows and full operating cost before funding an agent as a finished product. Method 13 Oct 2026 Use repository rules and a hold-out benchmark to review agent patches A change can pass the public test suite and still break a repository-specific compatibility contract. Google’s ADK case shows how maintainers can combine repository rules, a workload withheld from the agent and code review to decide which performance proposals are worth shipping. Method