Terranox AI (YC Winter 2026), Energy, Skoor 40
What they built
Terranox describes itself as "the first vertically integrated AI-powered uranium exploration company." [source] It says it finds high-grade uranium deposits in North America using AI trained on 70+ years of exploration outcomes. [source] The YC directory lists San Francisco, CA as its location, with Industrials and Energy as its industries. [source] The model architecture, the training data sources and any deposits found are not in the evidence.
Record
The Skoor is 40, based on n=4 signals, at confidence 0.8. All signals were observed on 2026-10-14.
| Signal | Weight | Detail | Source | |---|---|---|---| | Recent YC batch | 15 | Winter 2026 | [source] | | YC alumnus | 10 | "The first AI-powered uranium discovery company" | [source] | | Hiring | 10 | Hiring per the YC directory | [source] | | Small team | 5 | Team of 2 | [source] |
The YC directory lists the company as Active and at the Early stage. [source] No specific open roles are in the evidence. Founder names, backgrounds, funding, customers and code repositories are not in the evidence.
Inference: all four signals come from the YC directory. The Skoor therefore reflects YC admission and company self-description, with no independent check on the technology or the team.
Why now
The company states its mission as finding "the fuel we need to power the next century with nuclear energy." [source] Market data on uranium supply, demand or prices is not in the evidence.
Inference: the company's case rests on rising demand for nuclear fuel. This memo cannot verify that premise from the evidence.
Unknowns
- Who the two team members are, and whether either has geology or mining experience. Not in the evidence.
- Whether "vertically integrated" means the company holds land, claims or drilling rights. Not in the evidence.
- What "70+ years of exploration outcomes" consists of, and whether Terranox has exclusive access to it. Not in the evidence.
- Whether any prediction has been tested by drilling. Not in the evidence.
- Funding raised and roles being hired. Not in the evidence.
Questions for the room
- You say the model is trained on 70+ years of exploration outcomes. What are the datasets, who owns them, and could a competitor license the same data?
- You describe the company as vertically integrated. What land, claims or drilling rights do you hold today, and in which jurisdictions?
- You say you find high-grade deposits. Has any target generated by the model been drilled, and what did the cores show against your prediction?
What would move the Skoor
The Skoor would rise with founder backgrounds in geoscience or mining, held claims, or drill results matching model targets. It would fall if the training data is public and non-exclusive, or if no targets have been field-tested.