01 · LodestarOrbit
Where to point the satellite.
Meridian scores every area of interest by how much a new hyperspectral capture would change the answer. Then it tasks the capture over the best one.
How Lodestar works
The operating system for strategic resources
Intelligence that makes its own measurements.
Active learning decides what to measure next, at every scale. Coactive points hyperspectral satellites at the areas that matter, sends field teams to the samples that change the answer, and designs the lab experiments that settle it.
In the loop withPNNLNLRINLThorin Resources
San Juan Mountains, CO · real imagery and signatures · simulated tasking
Scroll
Why now
The United States imports its foundations.
In 2025 the U.S. was 100% import reliant for 13 critical minerals and more than half reliant for 20 more. China is a major source for 14 of those 33.
Finding more deposits is not the bottleneck. Knowing what a site can actually produce, and how cheaply, is.
Source: USGS Mineral Commodity Summaries 2026. Final 2025 List of Critical Minerals, excluding metallurgical coal and uranium.
13 100% import reliant20 more than 50%25 other critical minerals
One operating picture
Every source. One ground truth.
Orbit imagery, mineral signatures, terrain and legacy records, field samples, lab assays, and decisions, registered to the same ground. Every decision traces down through the layers to the measurements that made it.
- 01Tasked hyperspectral imageryLodestar
- 02Mineral signaturesLodestar
- 03Terrain and legacy recordsContext
- 04Field samplesTraverse
- 05Lab assaysCrucible
- 06Decision and next measurementMeridian
01 Imagery
02 Signatures03 Terrain
04 Samples
05 Assays
06 Decision
Active learning All the way down
02 · TraverseField
Which rock to pick up.
Inside the new scene, the same rule ranks every candidate site by expected information. The field team collects the few that change the ranking.
Watch it decide03 · CrucibleAutonomous lab
Which experiment to run.
In the lab, the same rule picks the next batch of conditions. Autonomous runs execute them and report back in days and weeks instead of months.
Inside Crucible04 · MeridianUpdate
Every result updates every level.
A lab result sharpens the site. The site sharpens the scene. The scene decides where the satellite looks next.
Meet MeridianOperate it
It proposes. You approve. It measures.
Meridian keeps a live queue of the next measurements worth making: satellite captures, field samples, and lab runs, ranked by how much each would change the decision. Captures and field work wait for a person. Lab runs execute on their own. Every result lands back in the picture and re-ranks the queue.
Meridian
San Juan district/Operating picture
Supervised autonomy
Day 1 · 06:00
Illustrative
LodestarMeridianTraverseCrucible
Illustrative interface running a simulation. Imagery, signature layer, and target positions are real (San Juan district, density peaks of the detected signatures). Rankings, timings, and results are simulated. Click a target to inspect it, or approve a field sample yourself.
Try it
Watch it decide.
This is the rule at field scale. The same rule points the satellite and designs the lab batch. One hidden deposit, forty-eight samples, the same blurry orbital prior. The grid walks the map in order. Active learning samples where the answer is least certain and most valuable, and updates after every result.
Grid sampling
Fixed planActive learning
Meridian-style policyBeliefHigh · dark is unmeasured
Illustrative simulation, not field data. Both panels share the same Gaussian-process model, the same noisy orbital prior, and the same budget of 48 samples; only the sampling rule differs. An upper-confidence rule stands in for Meridian's policy, which also weighs the cost of each measurement and the decision it serves.
On the ground
Real sites. Real rock. Real chemistry.
2,000+
Sites screened
Critical-mineral sites.
14
Priority targets
Ranked for field follow-up.
6
Permitted mines in the loop
Mapped below.
11
R&D sites
Research and development.
3
National laboratories
PNNL, NLR, and INL workstreams.
Ouray, COCrucible lab
Reno, NVCrucible lab
ArizonaPermitted mines
UtahPermitted mines
ColoradoPermitted mines
KentuckyPermitted mines
NLRGolden, CO
PNNLRichland, WA · CICERO
INLIdaho Falls, ID
Seattle
Washington, DC
Caddo LakeTX · LA · water monitoring
Screening coverage, western U.S.
Permitted mines in the loop, by state
Crucible labs
National laboratory workstream
Team
Two resources, one loop
What we point it at.
01 · ResourceMine loop pilot
Critical metals.
Domestic sites found, characterized, and qualified before anyone drills. The permitted Camp Bird and Revenue-Virginius mines are in the loop today.
Critical metals
02 · ResourceIn development
Water.
The resource under every other resource. Blooms, invasive species, and water quality, seen from orbit and confirmed in the water.
WaterCrucible · Ouray, CO · Reno, NV
A lab built for the model to drive.
Labs in Ouray, Colorado and Reno, Nevada, close to the rock they measure. XRF, ICP-OES, fire assay, pulverizing, drying, and separation, with Meridian choosing every run. Partner cells at PNNL and NLR extend it.
XRFQueued
ICP-OESQueued
PulverizerQueued
FurnaceQueued
Drying ovenQueued
CentrifugeQueuedPhotos: Ouray lab, 2026. Instrument readouts are illustrative.
Defense.Energy.Compute.
All of it starts with what a nation can find, verify, and run. Metals to build with and water to cool with sit upstream of every system that matters. Coactive works upstream.
Watch
Coactive, the AI for science platform.
Play · 4:38
Coactive Science · YouTube
Contact
Bring a resource and a decision.
Tell us the resource, the area, and the decision. We reply with whether Coactive can help and what a first screening would look like.
Or email us directly at info@coactive.science
Seattle, WA · Washington, DC · Ouray, CO · Reno, NV