The Loose Wheel Problem

Think about a race car. At 10 miles an hour, you might not notice a wheel that’s slightly out of alignment. The ride feels fine. Everything seems to work.

At 100, you feel the vibration. Something’s off, but you can still drive.

At 200, the wheels come off. You crash.

Every subtle gap in your architecture — every undocumented dependency, every governance shortcut, every fuzzy component boundary — is that loose wheel. At human speed, you can compensate. You notice things drifting. You course-correct. You ask someone and they fill in the gap.

At agentic speed, those same gaps are catastrophic. Agents don’t notice drift. They don’t self-correct based on intuition. They execute at velocity, and if the foundation has looseness in it, that looseness gets amplified with every increase in speed.

This is why I keep coming back to architecture. Not because it’s theoretically important, but because it’s practically essential at the speeds we’re starting to operate at. The tighter your foundation — your documentation, your interfaces, your validation loops — the faster you can safely go.

You don’t want to find out about the loose wheel at 200 miles an hour.