Blog

The region is bidding to lead the UK on AI: an AI Growth Alliance, a 2026–2029 skills strategy, ambitions for an AI Growth Zone. What changes for a business in Hampshire, and what to do now.

Part two of our GitHub Actions bill: one gate job per push, a €139 self-hosted runner behind two org-wide dials, and measured results, including what did not get faster.

AI makes production cheap faster than it makes coordination cheap, so the gains land on the few who review, decide and release. What actually fixes it.

A real GitHub Actions bill broken down: where the minutes go, why 48% of them are billing round-up, and the design that nearly halves the spend.

Requirements, design and build no longer happen in sequence — they evolve together, driven by AI agents. How AI-DLC differs, and where SystemDox fits.

Build a stateless, cache-safe multi-tenant API authorizer with Cognito Pre Token Generation v2 — and no per-request database lookups.

How application/problem+json gives API clients predictable, machine-readable error responses — and why it belongs in your Lambda handlers, not API Gateway.

How PuglieseWeb builds websites that live in your own GitHub repository: edit with SystemDox, auto-publish, full history, no vendor lock-in.

Dashboards answer yesterday's questions. Ad-hoc SQL on your event lake answers tomorrow's — user timelines, client debugging, and no runaway AWS bill.

Governance tools promise visibility into your data, but most small teams need to understand user behaviour, not Parquet file provenance.

How to build a career and a portfolio around Europe's accelerating shift away from US Big Tech towards open-source and European alternatives.

Publish CloudEvents-compliant business events through EventBridge to an S3 data lake, so analytics never couples to your operational database.

How we built an AI chatbot that lets business owners update their website by conversation — no CMS forms, no waiting for a developer.

A clear breakdown of how foundation models, skills, and retrieval-augmented generation work together to power modern AI applications.

How embedding models turn words into numbers that machines can compare, why that matters for search and AI, and what to consider when choosing one.

A practical explanation of how retrieval-augmented generation works, why vector databases matter, and when you should (and shouldn't) reach for them.

A practical guide to designing scalable, accurate AI agent systems with the Orchestrator-Specialist pattern for complex, multi-step automation.

Hard-won lessons from running AWS Lambda in production — layer path mistakes, silent auth failures, cold start surprises and the IAM traps.

The European stack for AI-powered apps — LLM inference, vector search, serverless and storage — for data sovereignty, lower cost and EU compliance.

Why tenancy is the one architectural choice that's genuinely painful to reverse — and the practical path from agency project to multi-tenant product.

A step-by-step walkthrough of how I use Claude Code to ship features faster — from spotting the problem to planning, implementing and testing.

A battle-tested stack for a modern site with AI chat, RAG, booking and admin — chosen for speed, maintainability and real-world running costs.

Architecture Decision Records only work if they stay current. Learn how to make ADRs a living part of your development workflow — not another stale wiki page.

Stop relying on manual code reviews to catch architecture violations. Learn how to automate guardrails with fitness tests that run on every pull request.

Practical event-driven architecture patterns using SQS, SNS, EventBridge, and DynamoDB Streams — with real-world trade-offs and when to use each.

Key insights from 444 technical writers, product managers, engineers and leaders on how teams build, measure and evolve their documentation.