Notes from the workshop

What we learned building things, written up while it was still fresh. No thought leadership, no listicles.

The difference between a model agreeing and a model being right

Ask a model whether your approach is sound and it will usually find a way to say yes. That tells you almost nothing, and we had to build a habit around it.

AIClaudeEngineering

AI · 6 min

08 Jul 2026

Build note: a translation layer for a supplier feed nobody could change

A retail client's stock feed arrived in a format from 2009, undocumented, from a supplier with no incentive to update it. The interesting engineering was all on our side of the wall.

Case studyEngineeringData

Case notes · 6 min

07 Jul 2026

The estimate that survives contact with a client

Most estimating advice is about getting the number right. The more useful skill is presenting a number that stays useful once reality disagrees with it.

ProcessDeliveryClients

Process · 5 min

06 Jul 2026

Testing the parts of an AI feature a user never sees

The visible output of a model-backed feature is the easiest part to check. The prompt assembly, the retries and the logging are where most defects actually live.

TestingAIQuality

Quality · 6 min

03 Jul 2026

What code review looks for that a linter cannot

Static analysis has absorbed most of the mechanical checks a reviewer used to do by hand. What is left is smaller and harder to automate away.

EngineeringQualityProcess

Engineering · 6 min

02 Jul 2026

Build note: a rota system that had to survive a strike

A scheduling tool for a care provider needed to keep working when the assumption behind every rota — that enough staff show up — stopped holding.

Case studyEngineeringProduct

Case notes · 6 min