AI has blind spots. So do good engineers.

AI has blind spots. So do good engineers.
Strong team. Battle-tested lead. Node.js stack. I trusted their calls and stayed at architecture level.
Then a single confused backoffice user broke everything.
Terrible UX first. Server crash second. We knew something was deeply wrong before it went down.
Root cause: async CRUD. Half-baked over-engineering. The kind of anti-pattern that seduced my confident tech lead — and that caught my PHP mind completely off-guard.
Years later, building with AI in the same stack.
As I run a basic test, I catch the same pattern. I recognise it immediately. AI went down the same rabbit hole — likely a community-learned pattern.
There’s no magic wand. Sometimes you move too fast, and you hit the wall. You learn.
Yesterday I reminded my team that async workflows brought complexity: I wanted none of it unless a specific use case would support it. Today I write that same set of instructions for my AI to get the right context.
The world does not change. It just moves faster.
What are the landmines in your stack — and have you briefed your AI around them?