Notes from the work.
What we learn building and running agentic systems, written down while it is still fresh. Most of it comes from our own testing platform and the applications it runs against.
-
Engineering 5 min read
We Use Our AI Testing Tool to Test Our AI Testing Tool
Aiqaramba production tests Aiqaramba staging. What the first run found, how we tell an app bug from an agent that got lost, and where our own testing still falls behind.
Read → -
Case study 5 min read
20 Bugs Sentry Didn't Flag: What AI Agents Found in MyLabelDesk
MyLabelDesk had Sentry in production and a quiet dashboard. We ran 25 AI agent test sessions across its 11-step workflow and found 20 bugs, including an upload that never left the browser and export buttons that did nothing.
Read →