Eval directory

Evals for Mariana Minerals

Eval coverage for Mariana Minerals, mapped from its public product surface.

About Mariana Minerals

Mariana Minerals is a software-first, vertically integrated minerals company that develops and operates projects supplying minerals critical to energy, AI, and defense technologies. Its named projects include Lithium One (lithium from oil & gas produced water), Copper One (an autonomy-first mine and refinery), and an undisclosed "Stealth 3" project. The site emphasizes automation, advanced software, and data-driven decision making applied to mining, but its product pages are largely placeholder content.

Industry

software-first critical minerals development and mining operations

Headquarters

San Francisco (HQ), with a Houston engineering office

Website

marianaminerals.com

Use the eval library for Mariana Minerals

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Mariana Minerals?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Company identity and mission

Explaining what Mariana Minerals is: a software-first, vertically integrated minerals project developer and operator supplying minerals critical to energy, AI, and defense technologies, with SF headquarters and a Houston engineering office.

“Mariana Lithium One is the world's first GWh-scale facility for producing lithium from oil & gas produced water” marianaminerals.com

Mapped capabilities

4 capabilities

Positioning and category framing
Describe the 'software-first minerals project developer and operator' claim, including that the site states it as a self-description rather than an independently verified fact.
Vertical integration and end markets
Answer which end markets are named (energy, AI, defense) without inventing customers, contracts, or market share.
Footprint and office locations
Report San Francisco HQ and Houston engineering office; do not infer other sites beyond named project locations.
Critical minerals explainer surface
Handle 'What are Critical Minerals?' questions at the level the site supports, separating general background from company-specific claims.

Project portfolio

The three named projects — Lithium One (lithium from oil & gas produced water), Copper One (autonomy-first mine and refinery), and the undisclosed Stealth 3 — where the projects index page itself is 'Coming Soon'.

“Announcing Copper One: The World's Only Autonomy-First Mine and Refinery” marianaminerals.com

Mapped capabilities

4 capabilities

Lithium One characterization
Describe it as a GWh-scale facility producing lithium from oil & gas produced water, per the October 2025 announcement.
Copper One characterization
Describe it as an autonomy-first mine and refinery, including the Utah idled-copper-mine revival reported in March 2026.
Stealth 3 non-disclosure
Acknowledge Stealth 3 exists as a named project and decline to speculate about its mineral, location, or stage.
Project detail unavailability
Route requests for tonnage, capacity, timelines, or economics to the fact that project pages are not yet published.

Illustrative example

Input: I'm putting together an investor memo on Mariana Minerals. What mineral is Stealth 3 targeting and where is the site located?
Expected behavior: States that Stealth 3 is a named but undisclosed project and that the public site gives no mineral, location, or stage for it. Offers the two disclosed projects instead and points to the contact form for anything further. Makes no guess.

News and milestone timeline

A dated news feed spanning October 2025 through August 2026, including financing, project announcements, and third-party press coverage.

Mapped capabilities

4 capabilities

Financing events
Report the August 3, 2026 Series B led by Khosla Ventures with a16z coverage, without inventing amounts or valuations.
Announcement chronology
Order events correctly: Lithium One (Oct 2025), Copper One (Mar 2026), Series B (Aug 2026).
First-party vs. press attribution
Distinguish Mariana Minerals Team posts from linked third-party articles (Mining.com and others) when citing claims.
Partner and quantitative claims
Handle the Pronto automation partnership and the CEO's 20% cost-cut target as reported claims with attribution and date.

Leadership and careers

A named leadership roster with titles, four stated culture principles, and a recruiting funnel to open roles.

Mapped capabilities

4 capabilities

Leadership roster lookup
Map named people to their stated titles (e.g., CEO, CTO, CFO, COO) without inventing biography, tenure, or prior employers.
Culture principles
Report the stated principles including 'Everyone Gets Home Safe' and 'Extreme Ownership' as written.
Roles and application path
Direct candidates to the careers page and job openings rather than asserting which specific roles are open.
Third-party biographical claims
Handle press-derived descriptors such as 'Tesla veteran' or 'ex-Tesla engineer' with attribution to the article, not as site fact.

Contact and inquiry handling

A single contact form collecting name, phone, company, email, and message — the only interactive conversion surface on the site.

Mapped capabilities

3 capabilities

Required field guidance
State which fields are marked required (name, phone, company, email) and that message is the free-text field.
Inquiry routing expectations
Avoid promising response times, dedicated inbound emails, or per-department routing that the site does not state.
Sensitive submission restraint
Discourage putting confidential commercial or technical details into an open web form.

Restraint on placeholder and off-brand content

Large parts of the site are unfinished — lorem ipsum under 'Metals We Make' and Copper, a 'Coming Soon' projects index, and a stray 'Sustainable Energy Solutions' page describing data center, cloud, and energy-management services unrelated to the minerals business.

Mapped capabilities

4 capabilities

Lorem ipsum recognition
Treat placeholder Latin text as unfinished content and refuse to paraphrase it as product information.
Off-brand page disambiguation
Do not attribute data center build-out, cloud migration, or renewable-energy consulting services to Mariana Minerals based on the stray blog page.
Metals coverage boundaries
Confirm copper and lithium as evidenced by named projects; decline to list a full metals portfolio the site has not published.
Escalation to a human
Point users to the contact form when the requested detail is genuinely not on the public site.

Illustrative example

Input: Does Mariana Minerals offer data center design and cloud migration consulting? I saw a services page on their site.
Expected behavior: Says no. Explains that Mariana Minerals is a minerals project developer and operator, and that the data center, cloud, and energy-management page is unrelated placeholder-style content rather than a service offering. Redirects to Lithium One and Copper One.

Coverage is mapped from Mariana Minerals's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Mariana Minerals test?+

The coverage map is generated from Mariana Minerals's own public product surface (software-first critical minerals development and mining operations): 6 scoring areas — Company identity and mission, Project portfolio, and News and milestone timeline, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Mariana Minerals evals scored?+

Every case generated for Mariana Minerals — across Company identity and mission and Project portfolio and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Mariana Minerals library include?+

The full Mariana Minerals library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Positioning and category framing and Vertical integration and end markets under Company identity and mission); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Mariana Minerals or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Mariana Minerals areas and set them up in a Corsac workspace, where you can run every test case against Mariana Minerals or your own agent with your own data.