Mozzarella Sticks

If you go to a restaurant and eat four fried mozzarella sticks, nobody thinks twice. It’s just an appetizer. But stand at your fridge a week later and eat four string cheeses - same ingredient, same quantity - and everyone in the house looks at you like you just committed a crime against humanity.

Context changes everything.

Last week I wrote about The Click - that moment when AI productivity locks in and everything moves. I also wrote about what happens next: spec debt. Simple requests for complex tasks. Velocity without discipline. Assumptions that compound until quality cracks.

That was my story. But it is also becoming everyone’s story.

A major tech company recently made headlines when AI-assisted code changes contributed to a string of outages across its retail platform and cloud infrastructure. The company’s own briefing described it as novel GenAI usage without fully established best practices and safeguards. Engineers were required to use AI tools aggressively - while the company simultaneously reduced headcount. More AI, more speed, fewer people to review the output.

The headlines blamed AI. But AI didn’t skip the code review. AI didn’t accept-all and commit without testing. AI didn’t decide that velocity mattered more than rigor. PEOPLE DID.

This is the mozzarella stick problem. The substance - AI-generated code - is the same whether it’s plated with discipline or grabbed from the fridge without thought. But when we skip the context - the specs, the reviews, the test-driven development, the 20% that humans are supposed to contribute - we don’t get to blame the cheese.

That metaphor might be a stretch, but I love mozzarella sticks. And #AI.

Spec debt is what happens when we stop putting our 20% in. When we treat AI like a finished product instead of a first draft. When we forget that the tool gives us the 80% and we still owe the rest.

The real question isn’t whether AI is ready. It’s whether PEOPLE are.