Quantom LLC — Three-Model Adversarial Analysis

Executive Summary

Three frontier AI models independently analyzed Quantom's technology and business potential, revealing consensus on key risks and opportunities while surfacing critical disagreements that illuminate the path forward.

What We Did

Independent Multi-Model Analysis

We fed identical documents from Quantom's data room into three independent frontier AI models:

Claude Opus 4.6 — Business strategy + integration thesis
Gemini Deep Think — Hard science skeptic, first-principles physics
GPT-5.2 Pro — 31 minutes of reasoning, constructive skeptic with citations

The models were NOT told what the others concluded. The disagreements reveal where the real insight lives.

Key Finding

Quantom Is A Hypothesis, Not A Product

"Very weak Raman," "not reproducible," "hypothesis—not a product."

— Quantom's own December 2025 gate memo

All three models converged on this assessment from Quantom's internal documents. The management team's honesty is a green flag — the question is whether the science can catch up to the vision.

3

AI Models

8

Patent Families

7

Weeks to P1 Deadline

$20M

SAFE Cap

Self-Seeding Nucleation (SSN) Process Illustration

Self-Seeding Nucleation (SSN): Electrochemical conversion of CO₂ into carbon nanomaterials

Where All Three Models Agree

When three frontier models independently converge, pay attention. These consensus findings represent the highest-confidence assessments.

Gate 1 Is Everything

Mechanism proof on gold is the only thing that matters right now. Everything else — market strategy, Tractor Beam, 40+ materials, carbon credits — is contingent on demonstrating reproducible off-seed propagation under controlled conditions.

Battery Additive Beachhead

The MWCNT battery additive strategy is the smartest commercial move. Real market (EV battery demand 4.5× by 2030), existing buyer demand, and CNTs occupy a genuine performance sweet spot.

Patent Deadline Crisis

April 3, 2026 (~7 weeks) is urgent. Non-provisional/PCT must be filed or priority is lost. The IP is only as strong as the enabling data behind it — currently "prophetic."

DSP Hidden Killer

Downstream purification could kill the economics. The cost of cleaning aqueous electrochemical output to battery-grade purity could dominate COGS.

Copper Cliff

If SSN only works on gold/noble substrates, the economics don't support battery additives. The "Copper Cliff" must be crossed. No copper = no beachhead.

AI Integration Natural Fit

SSN is a parameterizable control problem (seed + pulse recipe → morphology). Perfect for Bayesian optimization. But AI cannot break thermodynamics — it accelerates discovery OR falsification.

Where the Models Disagree

The disagreements between models surface nuances that no single analyst would catch. This is where the gold is.

SSN Scientific Plausibility

Opus:

"Unproven, needs Gate 1" — Business-balanced, defers to experimental results

Deep Think:

"Borders on thermodynamic impossibility" — Room-temp carbon adatom mobility near zero → amorphous soot. HER dominates. "If it works = Nobel Prize."

GPT-5.2 Pro:

"Not impossible but extraordinary" — Cites Nature Communications paper showing room-temp CO₂ → solid carbon. Three specific mechanisms identified.

Resolution: The ¹³CO₂ isotopic labeling test (GPT's recommendation) resolves this definitively. If ¹³C appears in the solid product, the carbon came from CO₂.

Tractor Beam Feasibility

Opus:

Significant engineering challenges noted but didn't deep-dive

Deep Think:

"Science fiction" — Paschen's Law, Earnshaw's theorem, monolayer graphene has zero bending rigidity

GPT-5.2 Pro:

"Electrostatic peeling is not crazy" — Cites ACS Central Science literature. Physics demos ≠ manufacturing

Recommendation: Treat Tractor Beam as a long-term R&D asset, not a near-term product. Focus all energy on SSN + battery beachhead.

Cross-Model Comparison Table

A comprehensive view of how each AI model assessed key aspects of the Quantom opportunity.

Finding	Opus 4.6	Deep Think	GPT-5.2 Pro
SSN Plausibility	"Unproven, needs Gate 1"	"Borders on thermodynamic impossibility"	"Not impossible but extraordinary" — cites Nature Comms precedent
Internal Doc Consistency	Not flagged	Not flagged	RED FLAG: Q1 2025 vs Dec 2025 timeline conflict
Off-seed Mechanisms	General risk	Adatom mobility wall	Three specific mechanisms + contamination as most likely
Tractor Beam	Noted challenges	"Science fiction"	"Not crazy" — cites ACS literature. Manufacturing ≠ physics demo.
Isotopic Labeling (¹³CO₂)	Not suggested	Not suggested	"Murders the contamination objection"
Investment Structure	Full acquisition	$30K Tranche A only	Option-to-acquire with milestone holdbacks
Tone	Business-balanced	Hard bear	Constructive skeptic

Five Highest-Impact Actions

Synthesized from all three models, ranked by leverage for the next 7 weeks.

1

Run Gate 1 With Critical Additions

Execute exactly as December 2025 memo specifies, plus GPT-5.2 Pro's recommendations:

¹³CO₂ isotopic labeling — proves carbon in product came from CO₂, not contamination
Mass balance / faradaic efficiency to solids — quantifies how much carbon becomes product vs gas/liquid byproducts

2

Pre-Register Controls Before Running Experiments

GPT laid out specific matched controls to prevent post-hoc goalpost shifting:

CO₂ vs N₂/Ar (same everything else)
Seeded vs no-seed (same pulsing)
Pulsed vs DC (same average current)
Seed smear control (handling without electrochemistry)
Electrically insulated seed (tests "conductive extension" hypothesis)

3

Triage Patent Portfolio Under April 3 Deadline

All three models flag this. Counsel-driven triage:

Must-convert: Core SSN (P1) + core apparatus
Nice-to-have: Speculative 40+ materials expansions

File narrow, defensible claims supported by real data — not broad claims that die on enablement challenges.

4

Demand Full Raw Data Transparency

GPT produced a detailed diligence request list. Key principle: if they won't show the full run history (not just best runs), assume selection bias.

5

Keep Tractor Beam as R&D, Not Critical Path

All three agree this is not ready for commercialization. It's an asset, not a product. Don't let it distract from the SSN → battery additive beachhead.

Decision Framework

A structured approach to investment decision-making based on gate criteria and milestone validation.

Cooperation / Transparency Gate

Complete raw datasets + full run history + full provisionals
Accept pre-registered controls + independent replication (ECL-style)

If NO: WALK (assume selection bias / story-first culture)

SSN Mechanism Proof on Gold

≥3σ off-seed metric exceeds control on ≥1 pulse recipe
Reproduced on ≥2 independent electrodes
Ring seeds outperform disks on off-seed metric
Strongly preferred: ¹³CO₂ incorporation evidence

If FAIL: One remediation loop max. If still fail → WALK

Copper Viability ("Copper Cliff")

Demonstrate propagation metric on copper substrate
Or scalable non-noble interface

If FAIL: Usually WALK for battery thesis (unless pivoting to premium/noble-substrate-only market)

Purity & DSP Economics

Document as-grown purity, DSP yields, waste costs
Show COGS not dominated by cleanup

Investment Recommendation: Option-to-acquire until Gates 1-3 clear, then full acquisition if AI integration potential validated

Critical Due Diligence Requests

GPT-5.2 Pro's comprehensive diligence package designed to kill false positives and surface IP/ownership problems.

SSN "Gate 1" Raw Data Package

Full run log for every SSN experiment to date (not just best runs). For each run, demand:

Unique run ID, date/time, operator
Cell schematic + photos, electrode geometry
Substrate prep SOP with timestamps
Electrolyte recipe (all chemicals, vendors, lots)
CO₂ delivery: mass-flow controller logs
Waveform files: exact pulse recipe
Raw Raman mapping package per pixel

If they won't show full run history, assume selection bias.

Critical Controls (Pre-registered)

Gas/control controls:

CO₂ vs N₂/Ar vs no gas flow
Cylinder + MFC only (no dry ice)

Seed/pulse controls:

DC vs pulsed at matched current
No seed + identical pulsing
Electrically insulated seed

The game-changer:

¹³CO₂ labeling — If ¹³C appears in solid via SIMS/isotope ratio/NMR, carbon came from CO₂. If not, contamination artifact (game over).

IP / Legal Diligence Package

Time-critical (7 weeks to April 3, 2026):

Copies of all 8 provisional applications (full text)
Exact April 3, 2026 family detail
Inventor list + signed assignment agreements
Lab notebooks / ELN exports supporting filings
All MTAs / SRAs / NDAs with universities
Concrete conversion plan with defensible claim scope

Critical: "Prophetic until validated" means current IP may be thin where it matters.

Tractor Beam Evidence Package

Request unedited videos of complete process:

Shear event → capture/hold → placement
Environment: air vs vacuum, humidity range
Sheet metrology after placement
Yield metrics: % successful single-layer pulls
Field parameters: voltage ranges, arcing events

If they can't produce this package, treat Tractor Beam as "concept art," not an asset.

The Feral Labs Thesis

"This analysis is itself the proof of concept. Three frontier AI models, working independently, delivering in hours what would take a traditional consulting engagement weeks and hundreds of thousands of dollars."

The disagreements between models surface nuances that no single analyst would catch. This is what Feral Labs does for portfolio companies. Not advice. Not decks. Infrastructure that thinks.

Murat — you and Mati have the full data room context. We have the analytical firepower. The combination is the point.

Next Step

Gate 1 at ECL. Let's make the science talk.