Insights

Page 2 of 7

A hand signs the last page of a printed specification with a fountain pen next to a closed laptop, in an office at dusk: signing off the spec before the agent generates code in Spec-Driven Development

Specs and agents: the specification is where you sign off

A railway signal operator rests a hand on a switch lever inside a signal box at dusk, facing a rail yard where the tracks diverge, as a metaphor for Jev, TypeSafe's decision model

Jev: the model that decides instead of writing. An in-depth analysis and ten use cases

An inspector sets a box aside from the sorting conveyor and examines it under a lamp at a side table at dusk, as a metaphor for the exception path of an AI-automated process

The exception path: what decides whether an AI automation works

A systems operator pauses at dusk at the glass partition separating a quiet control room from a live operations floor, as a metaphor for the boundary between an AI pilot and a company's live system

AI pilots don't die on the model: they die at the ERP boundary

A quality inspector at dusk checking a machined metal part against a reference gauge with a caliper, as a metaphor for the external oracle an AI-generated test needs

DevOps8 Sept 2026

AI-generated tests: a test written from the code doesn't test it, it describes it

An engineer signs at dusk for a pallet of sealed crates at an office loading bay and only lifts the lid of the top one, as a metaphor for the dependencies that arrive with AI-generated code

DevOps4 Sept 2026

AI-generated dependencies: the pull request line nobody argues about

Two engineers reviewing a large code change together on screen at dusk: one points at a specific line while the other takes notes, in a dimly lit technical office

DevOps14 Aug 2026

Reviewing AI-generated code: your review still expects an author

Two engineers in an office at dusk: one slides a single access card across the desk while keeping the rest of the stack, as a metaphor for minimum permissions in an AI agent

Prompt injection: your agent's security lives in the permissions, not in the prompt

A domain specialist reviewing and labelling evaluation cases for an AI system alongside an engineer, at dusk in a technical office

The golden set: how to build the cases that decide whether your AI works

An engineer reviewing a grid of evaluation criteria and a comparison between two versions of an AI agent across working screens

Evaluating AI agents: what "works better" actually means

An engineering team reviewing traces, evaluation results and cost metrics of an LLM system in production across several screens

LLMOps: the real path to production for an LLM system, and how to keep it alive

Leadership team at a mid-sized company reviewing a plan across a long table in early morning window light

Transformation2 Aug 2026

How to start with AI: the seven layers of maturity

Operations manager comparing a printed process report against figures on her laptop, annotating by hand

Transformation2 Aug 2026

AI value: choosing and measuring your first use case

Knowledge manager pulling a labelled folder from a well-ordered archive shelf

Transformation2 Aug 2026

Data for AI: the minimum viable preparation

Two colleagues reviewing and signing a printed policy document at a meeting table

Transformation2 Aug 2026

AI governance: the minimum required in 2026

Contact

Want to put these ideas to work in your team?

Let's talk about how to transform your organization with AI, DevOps and modern architectures.

What happens next

You write to us. Two lines are enough.
We reply within 24 working hours and set up a call.
On the call we tell you whether we can help and where to start.

Prefer to talk? Call us 682 384 429

Tell us what you want to solve

Tell us who you are and what you need.