Water exploration · United States

Drilling a well is a
$400,000 guess.

We built the model that removes the guess. It has read over a million drilled wells — and it marks the spot, and the depth, where your water is.

Request a pilot siting See the proof

01 — The problem

The invoice arrives whether you hit water or not.

Here is how a well gets drilled in 2026. You pick a spot — near the pump shed, where the driller has a feeling, or where a dowser’s stick dipped. The rig shows up. It charges by the foot: through the topsoil, through the clay, through eight hundred feet of maybe.

Then one of two things happens. Water — and nobody mentions the gamble again. Or sand, a trickle, two gallons a minute that won’t run a single sprinkler line — and you still owe the full amount. There is no refund for a dry hole. There is no insurance for it.

And it’s getting worse. Water tables are falling, the shallow easy water is gone, and wells that took 300 feet a generation ago now take 1,200 — at $300 to $700 per foot of maybe.

Two wells on the same land, under a kilometer apart, differ five-fold in water — in one case out of five. The ground is lumpy. The price of guessing wrong is not.

02 — What people try

Every way of picking the spot today is a version of asking around.

AThe dowser

A man with a bent rod, still hired across rural America. The USGS has tested dowsing for decades: it performs exactly at chance. In easy ground it “works” — because in easy ground, anywhere works.

BThe driller’s hunch

Experienced, sincere — and paid by the foot, hit or miss. His sample is the few hundred wells of one career. Nobody logs his misses.

CCopying the neighbor

The most reasonable bad idea in agriculture. One well in five differs five-fold from its neighbor. His water says little about your ground — and nothing about your depth.

DThe one-off consultant

A hydrogeologist samples one line across one field, for $10–30K. Better than a stick — but no consultant has seen more than a sliver of the state, and none will show you a track record.

03 — The model

Every well ever drilled already told us where the water is.

Every time anyone drills a well in America, the driller must file a report with their state: where, how deep, what layers the bit passed through, and how much water came out. California alone holds over a million of these reports in public archives — and even flew aircraft that scan aquifers like an X-ray — and published the data.

Almost nobody in the drilling industry has ever used any of it.

We trained a model on all of it. It has seen more drilling outcomes than every driller in California combined — and it learned what the ground looks like where wells hit big water, and where they come up dry. A hydrogeologist reviews every target before you see it. You get one thing:

Your siting report

Drill locationmarked on your parcel

Expected depthestimated to the layer

Expected flowrange, gallons per minute

Confidencestated, always

Legal right to pumpchecked against your basin

04 — The proof

Tested the only honest way: on the future.

We let the model learn only from wells drilled before 2018. Then it scored thousands of wells drilled after — wells it had never seen — and we checked its picks against what actually came out of the ground.

51%

hit strong water picking the usual way

95%

hit where our model said drill

hit where our model said don’t

Back-test on 23,772 California wells drilled 2018–2025, held out from training. “Strong” = ≥200 gallons/minute, the agricultural threshold. The model’s edge is largest where there are no nearby wells to copy — the exact places people hire dowsers.

Pilot program · 2026

Before you spend $400,000 on maybe — spend one email on knowing.

We’re running pilot sitings with a small number of growers, municipalities, and drilling contractors across the U.S. If a well is in your plans this year, we’d like to pick your spot.