5 May 2026|AI Engineering
AI-Augmented Engineering: Amnesia, Invented APIs and Team Knowledge Silos
Three practical AI engineering failure modes, with checks for session continuity, real API contracts and shared team knowledge.
Insights
Discoveries and developments from across the industry that support our engineering work: sovereign AI, agent grounding, compliance architecture and data strategy.
For our own build work, see Engineering →
27 April 2026|Agent Grounding
Team Chat Is a Knowledge Source, Not Ground Truth
How to turn team chat into traceable knowledge, with structured summaries, source links, permission checks and tests for stale answers.
24 April 2026|Agent Grounding
ADRs in the Agent Loop: Searchable Architecture Decision History
Use architecture decision records to give agents the reasoning behind a choice, with explicit approval, source links and searchable supersession.
22 April 2026|Agent Grounding
A Cache Hierarchy for Team Knowledge: Sources, Permissions and Freshness
Put a knowledge cache in front of team records without losing source ownership, access controls or a way to check freshness.
20 April 2026|Agent Grounding
Gotcha Files: Keep Engineering Bug History Useful
Record engineering gotchas with symptoms, evidence and tested fixes so people and agents can check past failures before repeating the investigation.
17 April 2026|Agent Grounding
Triple-Validated Retrieval: Evidence Checks for RAG Answers
Check stored claims, answer citations and uncited statements separately, with traceable evidence and explicit limits on what verification proves.
15 April 2026|Agent Grounding
Half-life decay for RAG: how to weight ageing knowledge
Time weighting can demote older retrieval results, but current answers also need valid timestamps, supersession and tests.
13 April 2026|Agent Grounding
AI Conversation Persistence: Save Each Turn, Then Summarise
Turn-by-turn capture needs a verified save boundary, replay-safe retries and summaries that point back to the original conversation.
10 April 2026|Agent Grounding
The Repo Matrix: A Living Document for Agent Grounding
Build a repo matrix from traceable code evidence, keep it current and make agents check existing contracts before proposing changes.
8 April 2026|Agent Grounding
MCP Tools for Team AI Memory: Discovery Without Blanket Access
Expose team AI memory through MCP tools with clear schemas, source references and separate write permissions, without replacing your existing API.
6 April 2026|AI Engineering
Test the extraction pipeline, not the model label
Compare extraction pipelines by accepted records, validation failures and retry costs before choosing a smaller model for speed.
3 April 2026|Sovereign AI
The Single-Stream Benchmark Trap: How to Evaluate AI Inference Hardware
Evaluate AI inference hardware at realistic concurrency, with throughput, latency and output quality measured together.
1 April 2026|Sovereign AI
What to Test Before Spending £4,000 on Local AI Hardware
Assess local AI hardware through model fit, useful capacity, data boundaries and ownership costs before committing a £4,000 budget.
30 March 2026|AI Engineering
Heavy, Medium, Light: Routing LLM Workloads by Task
Route LLM workloads by required quality, response time and failure cost, with tested escalation rather than assumptions about model size.
27 March 2026|AI Engineering
Reasoning Effort: Test the Token Budget Before Buying More Capacity
Reasoning effort changes how gpt-oss approaches a task. Test output budgets and accepted records before claiming a throughput gain.
25 March 2026|AI Engineering
AI Procurement: Three Questions Beyond 'AI-Powered'
Evaluate an AI product through answer provenance, failure handling and data location, with evidence to request before procurement.
23 March 2026|Compliance
Offline-First State Management: Building Data Continuity for Compliance
Design offline compliance records with explicit save states, traceable corrections and tested server acceptance, rather than a single sync indicator.
20 March 2026|Mobile & Offline
Microservices for Scalable Field Ops
The challenge of managing field operations often involves fragmented data and disconnected systems. Operational Directors and CTOs grapple with the problem of
18 March 2026|Sovereign AI
Local Inference: Securing Data Flow Without False Guarantees
Local inference changes where processing happens. Protected records, access controls and tested audit trails establish what happened to the data.
13 March 2026|Compliance
Automated Compliance Workflows: Rules, Approvals and Evidence
Automated compliance workflows need versioned rules, accountable approvals and retrievable evidence. AI should assist within those controls.
11 March 2026|Compliance
Mobile Safety Auditing: Building a Traceable Record
Design mobile safety audits around evidence, corrective actions and review, while keeping England's building safety golden thread duties in scope.
9 March 2026|Data Strategy
API-First Design: Make the Golden Thread Testable
Design API contracts around traceable field records, explicit save outcomes and tested recovery, rather than promises of automatic data continuity.
6 March 2026|Compliance
GDPR Compliance in Offline Apps: Retention, Erasure and Audit Records
Offline GDPR compliance needs explicit retention rules, deletion acknowledgements and tests that stop old devices restoring erased records.
4 March 2026|Data Strategy
Legacy SQL to NoSQL Migration Strategies for Smart Logging Systems
Plan SQL to NoSQL migration around record meaning, repeatable change capture and tested rollback, rather than a promise of greater scale.
2 March 2026|Mobile & Offline
React Native vs Capacitor: Preserving the Golden Thread
Assess React Native and Capacitor through record identity, correction history and evidence handover, rather than promises of automatic compliance.
27 February 2026|Data Strategy
Serverless Architecture for Spiky Smart Logging Workloads
Serverless logging needs bounded concurrency, safe retries and tested recovery. Automatic scaling alone does not prove data continuity.
25 February 2026|Mobile & Offline
Battery Optimisation for Field Service Apps: Measure the Work
Measure battery use against real field tasks, then tune location, transfers and background work without hiding delays or weakening records.
23 February 2026|Data Strategy
Real-Time Collaboration Constraints: Keep Changes Traceable
Real-time collaboration needs explicit conflict rules, clear save states and protected history so shared field records remain traceable.
18 February 2026|Mobile & Offline
React Native vs Capacitor for Enterprise Smart Logging
Compare React Native and Capacitor through local storage, native integrations and recovery tests, rather than promises about data continuity.
16 February 2026|Compliance
Geo-Fencing for Asset Tracking: Evidence Before Compliance
Use geofencing for asset tracking without confusing a device alert with continuous location evidence or proof of building safety compliance.
13 February 2026|Mobile & Offline
Field App User Experience: Keeping Field Records Understandable
Design field app user experience around clear labels, accessible errors and truthful save status, with ownership and recovery tested before release.
11 February 2026|Mobile & Offline
Biometric Authentication: What It Proves in Field Operations
Biometric authentication can support field accountability, but permissions, protected logs and offline rules must establish what happened.
9 February 2026|Data Strategy
Microservices for Scalable Field Operations: Establishing Your Golden Thread
Microservices can separate field workloads, but traceable records depend on service boundaries, safe retries and explicit ownership.
6 February 2026|Mobile & Offline
Vector Databases on Edge Devices: Search Is Not the Record
Use vector search to find related field records offline, while keeping original observations, permissions and synchronisation separate.
30 January 2026|Mobile & Offline
Offline-First State Management: Redux and MobX for Field Apps
Separate screen state from saved records when choosing Redux or MobX for offline-first field apps, with explicit recovery and synchronisation rules.
28 January 2026|Mobile & Offline
Cross-Platform UI Frameworks: Keep the Data Contract Independent
Choose cross-platform UI frameworks around a shared data contract, with explicit record meanings, portable exports and tested upgrade behaviour.
26 January 2026|Mobile & Offline
Testing Strategies for Offline Scenarios: Ensuring Data Continuity
Test offline apps through interrupted saves, lost responses and conflicting edits, with evidence that records survive and synchronise correctly.
23 January 2026|Data Strategy
Why Gut Feeling Needs Testing: Data-Driven AI in Enterprise Operations
Test operational AI against expert judgement and simple rules, with traceable data, explicit limits and evidence from the work it supports.
23 January 2026|Mobile & Offline
Data Science for Offline-First Enterprise Apps
Make field predictions useful offline: validate inputs, test the data held on the device and measure the decision rather than model accuracy alone.
22 January 2026|Data Strategy
Scalable Data Architecture: Measure Capacity Before Changing the Design
Design scalable data architecture around measured demand, queue recovery and database behaviour, rather than assuming more services will solve capacity.
17 January 2026|Compliance
Enterprise App Security: Why Hidden URLs Are Not Access Control
Hidden buttons and obscure URLs do not enforce permissions. Put access checks on the server and test what each user must be refused.
12 January 2026|Data Strategy
Dashboard Fatigue: Designing for Action over Observation
Design actionable business intelligence around decisions, named owners and reliable data, with alerts reserved for work that needs a response.
2 January 2026|Mobile & Offline
Native vs Hybrid ROI: Costing a Capacitor App
Compare cross-platform app development cost through shared code, native integration, testing and upgrades, rather than assumed savings.
11 March 2025|Compliance
Digital Twins and the Golden Thread: Evidence Before Prediction
Use digital twins for building safety decisions with traceable records, tested models and explicit limits on prediction and automated action.
2 March 2025|Mobile & Offline
Automated Decision Trees for Offline-First Applications
How to design offline decision trees with explicit rules, local inputs, versioned policies and safe synchronisation when connections return.
25 February 2025|Data Strategy
Database Normalisation: The Hidden Cost of Bad Data Hygiene
Treat database normalisation as a question of record meaning, enforceable relationships and correction work, rather than a guarantee of clean data.
23 February 2025|Compliance
Temporal Root Cause Analysis with Immutable Event Logs
Use protected event histories to investigate incidents, test causal explanations and choose leading indicators without promising predictive safety.
18 February 2025|Data Strategy
Technical Debt Management: Fix the Data Rules Before the Rewrite
Manage technical debt in offline field apps with explicit data rules, tested conflict handling and staged migrations, rather than a wholesale rewrite.
16 February 2025|Sovereign AI
Federated Learning on Edge Devices: Training Local LLMs Privately
Federated learning keeps raw training data on devices, but private mobile LLM training still needs leakage controls and a measured deployment plan.
13 February 2025|Mobile & Offline
Field Service Management: Scale the Data, Not Just the User Count
Size field service systems for queued work, reconnect bursts and conflicting updates, rather than relying on registered user counts.
13 February 2025|Sovereign AI
Optimising Local LLM Inference Latency for Field Applications
Measure local LLM inference latency against the field task, then test model compression, native integration and offline behaviour on real devices.
11 February 2025|Data Strategy
Single Source of Truth: Define Ownership Before the API
Build a single source of truth around data ownership, explicit API contracts and tested rules for stale copies and conflicting changes.
9 February 2025|Data Strategy
Data Hygiene: Make Business Data Fit for Its Purpose
Define useful data quality rules, find errors at source and keep corrections traceable across business systems and offline field apps.
9 February 2025|Mobile & Offline
Legacy System Modernisation: When Mobile Wrappers Leave Field Teams Stranded
Modernise field workflows by testing offline work and API boundaries, rather than assuming a mobile wrapper fixes a legacy application.
6 February 2025|Data Strategy
Data Architecture in Construction Projects: Control the Information Handover
Construction data architecture needs clear ownership, revision control and tested handovers, including what happens when site devices work offline.
3 February 2025|Mobile & Offline
Offline-First: Why Gut Feeling Fails in Enterprise Operations
Offline-first field apps need clear rules for local saves, retries and conflicting edits. Define what each action means before choosing the tools.
2 February 2025|Data Strategy
Database Normalisation for Cost-Effective Cloud Infrastructure
Database normalisation reduces duplication, but savings need evidence. Separate data integrity from query performance before changing your cloud database.
1 February 2025|Mobile & Offline
Securing Mobile Data: A Logic-First Approach
Secure offline mobile data with explicit access rules, protected local storage and evidence that controls work when devices lose connectivity.
26 January 2025|Data Strategy
Taming the Cloud: Controlling Infrastructure Costs
Cloud cost optimisation starts with ownership and measured demand. Use allocation, scaling and budget controls without promising automatic savings.
Nothing matches that yet. Try another word or topic.