Chunking

Chunking is breaking AI work into pieces small enough that each piece’s output can be verified in under five minutes — the five-minute rule. Size the chunk by verification time, not by how much work it feels like: a chunk is right when you can confirm it’s correct before the next one starts, so errors get caught at the source instead of propagating through everything built on top of them.

Why it matters: The over-prompt pattern — handing AI a huge job in one huge prompt and getting output that looks impressive for thirty seconds and can’t be trusted anywhere. The failure isn’t the AI; it’s a two-hundred-hour job assigned with no decomposition and no checkpoints, which would break a human contractor the same way. (Honest lineage: decomposing work into verifiable pieces is older than software — the five-minute verification target for AI work is the new part.)