Notes from the workshop

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

Running two coding agents on the same codebase

We tried assigning parallel tickets to two agents at once on the same repository. The bottleneck was not the agents, it was the branch strategy.

AIEngineeringDelivery

AI · 6 min

10 Jun 2026

Build note: a permissions model that outgrew its first design

A simple role-based system worked for a year, then a client asked for an exception that broke the model. What we replaced it with, and why we waited.

Case studyArchitectureEngineering

Case notes · 7 min

09 Jun 2026

The cost of a flaky test suite

A test that fails intermittently is worse than no test at all, because it trains the team to ignore red. We put a number on what that costs.

TestingQualityEngineering

Quality · 6 min

08 Jun 2026

How an agent changed our onboarding week

New engineers used to spend their first week reading. We rebuilt the week around a coding agent as a guided tour of the codebase, with mixed but useful results.

AIProcessEngineering

AI · 6 min

05 Jun 2026

What a staging environment is actually for

Staging environments quietly decay into a slightly different production. Getting the value back means being explicit about what they exist to catch.

EngineeringQualityProcess

Engineering · 5 min

04 Jun 2026

Letting a model write the first test, not the first fix

When a bug report comes in, we now ask the model to reproduce it as a failing test before anyone — human or model — is allowed to propose a fix.

AITestingQuality

AI · 6 min