#12 in an irregular series
A geologist and seven companions spent five weeks driving across the Arabian desert. They were looking for signs that anyone had lived by its ancient lakes.
And they found the lakes. But no sign anyone had lived there.
Given this underwhelm, it is entirely understandable that the hypegeist missed it. But it pays those of us moving AI agents from pilots into production to understand the paper’s stroke of genius.
For years, human migration theories assumed people followed the water. Until now, nobody had properly scoured the interior for people. And when the team searched they indeed found lake beds, snail shells and ostrich eggshell. But no obvious indication of human habitation. Why?
The missing ingredient was stone. Back then, people could carry stone a long way. But the further it had to go, the less of it that arrived - enough for a visit, not a lifetime.
At well-studied Neanderthal sites in Europe, most stone came from within about 5 km. But these lakes were over 200 km from any known outcrop of usable stone.
And yet the previous theories of human migration had only mapped the water, not the stone as well.
AI pilots also tend to go where the water is: a clear use case, a keen sponsor. And that’s often the right place to start. But a pilot is a visit. Many run on stone carried in by hand: someone exports the data, handles odd cases, chases access requests. And the carrying rarely shows up in the results.
When the project ends, who keeps carrying the stone?
Early humans couldn’t move an outcrop to the water. But fortunately, a contemporary business can.
During the pilot, identify what and who are doing the carrying. Each fix by hand can show what’s missing. Pass those fixes and their reasons to someone with time and budget to change the source: a feed instead of an export, the cause instead of each case.
Then there’s less to carry.
Because there is a lot of lovely water out there. But Neanderthals remind us to make sure the sites we set up have everything we need for the long run.
