A Four-Tier Trust Model for Evaluating Claude Code Skills
Popular Claude Code skills can make output worse. Use a four-tier trust model and a build-vs-borrow filter to pick skills your team can rely on. Read the guide
AI as a mentor, the Socrates way
Using AI as a mentor: how one Zartis engineer built a Socratic Claude skill that questions and challenges him while he studies. Read how to build yours.
Stop Maintaining Your Service Catalog by Hand
Your pipeline already knows who shipped what, and where. See how we built an automated service catalog that stays current. Read the full story.
QUERY: The HTTP Verb AI Agents Needed
Learn how the HTTP QUERY method separates safe reads from writes and tightens AI agent permissions. See what it means for your architecture.
Machine-Readable Marking of AI Content: What Article 50 Compliance Looks Like in the Pipeline
AI Act marking fails in the middle of the pipeline. How C2PA, watermarking and preservation actually work, and what to build for Article 50.
You are a deployer: the AI Act transparency rules for your own content
Article 50 makes ordinary companies responsible for labelling AI content. What deployer obligations cover, and where your supply chain leaves a gap.
How To Evaluate And Select A Trusted AI Technology Partner For Highly Regulated Sectors
A CFO’s checklist for choosing a trusted AI technology partner in financial services, healthcare and energy. Six questions to ask before you sign.
The Agentic AI Implementation Guide for Banking: Governance Architecture and Delivery
How banks are building agentic AI governance from day one, covering SR 26-2, the EU AI Act, and FCA rules. See what CTOs need in place
Evaluating Generative AI Development Services for Custom Enterprise Applications
How to evaluate generative AI development services for custom enterprise applications: a three-part framework covering technical depth, delivery and governance.
Evals and Observability: The Two Things Every Enterprise Asks About Before an AI System Goes Live
Why evals and observability are now baseline requirements for production AI agents, and how to build both into your delivery pipeline. A technical guide
Context Engineering: The Discipline That Decides Whether Your AI Agent Actually Works
AI agents fail on bad context, not bad prompts. Learn the five disciplines of context engineering, from retrieval design to CLAUDE.md, with real examples.
Your AI Pipeline Is Fast. Your Review Gate Needs to Catch Up.
AI speeds up code generation, but naive review gates create a new bottleneck. Learn how to design risk-tiered, async human-in-the-loop checkpoints that scale.