Vancouver, BC  ·  Deep-Tech Studio

The only venture designing alloys
from the junk already in orbit.
Not for it. From it.

Argead builds the translation layer between orbital debris composition and printable refractory high-entropy alloys — spectroscopy in, fabrication spec out, computed at the point of capture.

ONE ALLOY. ONE ORBIT. AT A TIME.
The Empty Square

Two industries. Neither has claimed the mapping.

Alloy-design AI assumes clean, curated feedstock — you can order pure elements. Debris-removal ventures treat every derelict object as trash to be deorbited and burned. In-space manufacturing launches its feedstock from Earth at thousands of dollars per kilogram.

Nobody has built the function that turns an unknown, contaminated, mixed-metal object in LEO into a viable high-entropy alloy composition and a fabrication spec. That mapping requires fluency in both domains at once — and the two industries don't talk to each other.

Aerospace Doctrine
“The best part is no part. Deorbit the junk.”
Argead
“The best part is the part that's already in orbit.”

We aren't contradicting the doctrine of eliminating mass. We're extending it one step further: mass in LEO has already been paid for. Roughly $2,500 per kilogram of launch cost, sitting there as inventory.

The Translation Layer

Spectroscopy in. Fabrication spec out.

Onboard spectroscopy reads the elemental fingerprint of a captured object. A generative surrogate model, running on edge GPUs, maps that fingerprint and its orbital environment to a refractory high-entropy alloy composition and a laser-sinter fabrication spec. This mapping — (spectroscopy input, orbital fingerprint) → (composition, fab spec) — is the IP.

Live model schematic — cycling three internal candidates [COMP] Computational
Built For

Environments, not logos.

Reentry leading edges. Fusion plasma-facing walls. Turbine hot sections. LEO hardware that can't come home for repairs.

The organizations that operate in these environments know who they are — and they know the difference between a validated alloy and a prediction. We're building toward the former and labeling the latter.

Stage Zero

Three candidates. None validated. That's the point.

These compositions were produced by a variational-autoencoder framework over a curated refractory HEA corpus. They exist as computational predictions and nothing more — and we label them that way, structurally, in every output we produce.

AV-001
Hf₁₅Mo₁₀Nb₂₀Ta₂₅Ti₁₀W₁₀Zr₁₀
[COMP] Computational
AV-002
Mo₁₅Nb₂₅Ta₃₀W₂₀Zr₁₀
[COMP] Computational
AV-003
Hf₁₀Nb₂₀Ta₂₅Ti₁₅W₂₀Zr₁₀
[COMP] Computational
Not yet arc-melt validated. Every prediction on this page carries this disclosure until furnace data exists. Honest positioning is not a caveat here — it's the product spec. The evaluators we build for read overclaiming as a defect.

Full prediction ledger →

Interrogate the corpus yourself →

The Log

What we did, what we know, how we know it.

Every entry carries an evidence-basis tag: [EXP] experimental, [COMP] computational, [LIT-INF] literature-inferred. If a claim can't carry a tag, it doesn't ship.

Channels

Three reasons to write.

Research — collaboration on refractory HEA design, validation, or the debris-to-feedstock mapping.

Programs — agency and prime program discussions. We'll tell you plainly what's validated and what isn't.

Corpus access — a login to interrogate the corpus yourself.