The other day I was wondering to myself, "how did farmers figure out that Gilroy was the ideal place to grow garlic?"

They didn't!

Presumably over the last hundred or so years lots of farmers randomly tried growing garlic in lots of random places.

The likelihood they tried it is inversely tied to how much people expect it to fail in that location based on folk knowledge of previous failures.

If it didn't do well in that environment, after the season they won't do it again (and maybe will share that knowledge that it failed, directly or indirectly e.g. "Everyone knows that Steve tried garlic five years ago and it bankrupted him and he left the town in shame.").

After each season a farmer will decide whether to keep the same crop or swap to a different one.

If it went well, they're unlikely to switch.

But if there are other ones they think might do better they might switch.

Over time, the garlic farm 'grows' in the place it's best for; the other places swap to a different crop quickly.

A distributed percolating decision making process with no top-down component, that gives a high-quality answer that looks like it was made intelligently by a single actor but was actually decided by the swarms' collective intelligence.