AIsopach

Subsurface intelligence · AI Reservoir

A century of drilling left 434 billion barrels behind. We find them at the speed of AI.

Logs, scout tickets and production history from 4.85 million American wells, read at machine scale. AIsopach finds the unrecovered hydrocarbons a century of operators drove past — and ranks them, title-cleared, before anyone else knows they're there.

The land & title suite is live today. The reservoir engine is in production on our own deal flow.

Isopach · interval rank · section sweep● Live
08
Flagged intervals
46
Logs read, this section
624Bbbl
Discovered in US reservoirs
Conventional oil found to date
434Bbbl
Still in the ground
Left after primary and secondary recovery
4.85M
Wells drilled in the US
Every one of them logged something
770K
Marginal wells on production
88–97% of onshore conventional verticals

01 · The gap

The barrels were never lost. Nobody could read fast enough.

Two thirds of every barrel ever discovered in the United States is still down there — logged, measured, written down, and walked past because no team of geologists could re-read a century of archives. Machines can, and now do.

01

Logged once, in 1958.

Vintage raster logs that were never digitized, never normalized across vendors, and never compared with what the well two sections over actually produced.

02

Cased off, not condemned.

Pay above and below the target gets cased off on the way to the zone somebody was paid to drill. The log still saw it. The report never mentioned it.

03

Evaluated by hand, one section at a time.

A geologist works through sections at human speed. There are hundreds of thousands of them, and the ones nobody got to are exactly where the value hides.

An interval the original operator cased off. The engine flags it because its signature matches intervals that produced in the same geology — illustrative, not a specific well.

02 · The engine

Four systems. One ranked list.

AIsopach is not a dashboard. It is an engine that ends in a ranked, title-cleared, economically screened target list — and it runs continuously, not per project.

E-01In production

Reads what the archive already knows

Raster and digital logs across vintages, vendors and units, pulled into one comparable signal. The engine flags intervals whose response matches what actually produced elsewhere — not what the original operator happened to be looking for.

Four vintages of the same formation, read into one comparable signal.
E-02In production

Trained on outcomes, not on theory

Production history is the label. The engine learns which intervals produced, in which geology, under which completion — so the ranking sharpens every time a new well reports.

Sections scored inside one play, best three carried forward.
E-03Live today

The landman suite, already running

Runsheets, chains of title and net-mineral-acre positions assembled out of county records automatically. This is the part that is built and in use today: title runsheets go from days to 24 hours at high accuracy, and land acquisition costs fall 70% against contract-landman rates.

A chain of title assembled automatically, ending in a net-mineral-acre position.
E-04In production

Ranked targets, with the math attached

Each target arrives with interval economics, resolved ownership and a recommended structure, then goes to Fifth Meridian Energy to acquire — royalty, minerals, non-op or operated.

Illustrative. Real ranks, sections and economics are released under NDA.
01

Ingest

Logs, records, production

02

Read

Normalize and flag intervals

03

Rank

Score against produced analogs

04

Title

Live

Runsheet, chain, NMA

05

Acquire

Hand off to Fifth Meridian

03 · The market

The money is moving toward this, not away.

The barrels are already booked as discovered. The land work is already being paid for by hand. The only thing that has changed is that reading all of it at once is now possible.

The size of what was left

137Bbbl
Recoverable with CO₂-EOR alone
Before any new interval is added
50%+
Of those reservoirs in TX & the Gulf Coast
Where the land records are deepest

Upstream digital & AI spend

Annual market, 2025 to 2030 — against a projected $500B of cumulative value to E&P companies over the same window.

What land work costs today

$575/day
Fully weighted contract landman
Industry range $500–650
$5B/yr
Cost of wrong ownership calculations
Industry-wide, from manual interest math
75%
Title timelines cut by connected workflow
Before any AI is applied at all

04 · Disclosure

The method is the asset. This page is what we say out loud.

We would rather be underestimated in public than copied. Everything below is available in detail, under NDA, to a named firm.

The feature set and the ranking model

Not published. Not licensed. Not described beyond this page.

The training corpus and how outcomes are labeled

Not published, and not part of any evaluation copy.

Which basins and which sections are currently ranked

Named under NDA only, in the data room.

What you do get, on an executed NDA: the technical brief, the current basin list, the land & title benchmarks measured against contract-landman rates, and the Fifth Meridian structure the engine feeds.

Access

Ask for the room, not the brochure.

Requests from named firms get a reply. Access includes the technical brief, the current basin list, the land & title benchmarks, and how Fifth Meridian Energy is structured around the engine.

48hr
Typical reply time
NDA
Before anything specific