Skip to main content
Navigated to Ethics & Transparency
    Ethics & Transparency

    How We Operate

    Helical Systems LLC, doing business as Vrnl ("we," "us," or "our"), believes in full transparency. Here's what we do with your data, how we present information, and what commitments we make to accuracy and the environment.

    The Trade-Off

    Vrnl uses AI to analyze environmental data. AI consumes electricity and water. We've estimated this trade-off to the best of our ability. Here's our reasoning for why we believe the outcome is net positive.

    Our estimates vs. published research on downstream impacts

    One Vrnl Report (our estimate)

    • Energy: ~0.5–1 Wh
    • CO₂: ~0.06–0.12g
    • Water: ~0.5–1 mL (10–20 drops)

    Equivalent to watching TV for 18–36 seconds

    Potential Downstream Impacts

    • ER visit: 20–80 kg CO₂*
    • Annual chronic care: dozens to hundreds kg CO₂*
    • Environmental remediation: significant CO₂ + resources*

    *Ranges based on healthcare emissions research. Varies significantly by region, facility, and treatment.

    Our reasoning: If even a small fraction of reports help people avoid environmental health exposures, the CO₂ spent on information is vastly outweighed by the CO₂ of the healthcare, remediation, and resource waste that would otherwise occur.

    This is a logical argument, not a measured outcome. We cannot prove causation between information and prevented healthcare interventions.

    Our Footprint: The Numbers

    We estimate our AI footprint using published research from Google, our infrastructure provider.

    What We Know vs. What We Estimate

    What We Know

    Per-query AI energy consumption (Google's disclosure). Our infrastructure provider's efficiency metrics.

    What We Estimate

    Queries per report (~2–4). Total per-report footprint (derived from above).

    What We Believe But Cannot Prove

    That informed decisions prevent some healthcare interventions. That prevention is less resource-intensive than treatment.

    Per AI Query (Google Gemini, August 2025)

    0.24

    Wh energy

    0.03

    g CO₂e

    0.26

    mL water

    Our estimate: Each Vrnl report uses approximately 2–4 AI queries for analysis and summary generation.

    Total per report (estimated): ~0.5–1 Wh energy, ~0.06–0.12g CO₂, ~0.5–1 mL water

    These are estimates based on published methodology. Actual figures vary by query complexity and infrastructure load. We acknowledge margins of error.

    Where Our AI Runs

    Vrnl's AI runs on Google Cloud infrastructure, which has made public commitments to efficiency and renewable energy.

    33× More Efficient

    Google reported a 33× improvement in energy efficiency per AI prompt between May 2024 and May 2025.

    1.09 PUE

    Google's data centers operate at 1.09 Power Usage Effectiveness—among the most efficient in the industry (closer to 1.0 is better).

    Carbon-Free Target

    Google Cloud is working toward 24/7 carbon-free energy for all operations, not just annual offsets.

    120% Water Replenishment

    Google's goal is to replenish 120% of the freshwater consumed by their data centers.

    Source: Google Environmental Report 2025

    The Cost of Not Knowing

    Information has an environmental cost. Ignorance has a larger one. This section presents research-backed context, not claims we can prove about our specific impact.

    Healthcare's Carbon Footprint

    The global healthcare system generates approximately 4.4% of worldwide greenhouse gas emissions (Health Care Without Harm). Preventable health conditions from environmental exposures contribute to this burden.

    Environmental Health Economics

    Proximity to high-pollution sources is associated with increased respiratory and cardiovascular healthcare costs. Early awareness allows buyers to factor these potential health costs into their decisions—or choose differently.

    Before vs. After

    Remediation after purchase is significantly more resource-intensive than informed avoidance or negotiated disclosure before purchase. Prevention is cheaper in carbon and dollars.

    The specific ratio varies widely by situation—we cannot provide a universal multiplier.

    How We Minimize Impact

    We don't use live AI for everything. We engineer responsibly:

    Smart Caching

    Environmental hazard data is stored efficiently. We don't re-query AI for information we've already mapped.

    Efficient Queries

    Our queries are optimized to be as lightweight as possible, reducing load on data centers.

    Pre-Computed Analysis

    Much of our environmental data is pre-processed and indexed. AI is used for synthesis, not raw data retrieval.

    Continuous Optimization

    We regularly audit our AI usage patterns and refine prompts to reduce unnecessary computation.

    Our Pledge: Specific Commitments

    1% Giving Pledge

    By the end of 2026, Vrnl will donate 1% of gross report revenue to certified environmental remediation organizations.

    Annual Reporting

    We will publish our first giving report by the end of 2026 and update it annually on this page. Partners will be listed once finalized.

    Footprint Transparency

    Starting in 2026, we will publish annual estimates of our total AI energy consumption using the methodology described above.

    We acknowledge these are commitments, not achievements. We'll update this page as we meet them—or explain why we haven't.

    How We Present Information

    We take editorial responsibility seriously. Our reports follow specific guidelines:

    Concern Levels, Not "Risk": We use "lower concern," "moderate concern," "elevated concern" instead of "low risk," "high risk." Risk implies probability we cannot calculate.
    Context Over Alarm: We always provide EPA threshold context. Detection is not danger—we clarify when levels are below regulatory thresholds.
    Data Vintage Transparency: Every report section shows when the underlying data was collected or last updated.
    Qualified Language: When we find no data, we say "We found no [X] based on available data"—not "There is no [X]."
    No Fearmongering: We design for informed decisions, not fear. Our internal test: Would an environmental scientist read this and call it fair?

    Our Content Standards

    AI-Assisted, Human-Reviewed: Blog posts and educational content are drafted with AI assistance and reviewed by human editors. We disclose this on every article.
    Corrections Policy: If we publish something inaccurate, we correct it promptly with a visible update notice.
    No Sponsored Content: We do not accept payment for blog coverage or data placement. Rankings reflect data, not commercial relationships.
    Source Attribution: We cite specific data sources with dates in our content.

    How We Make Money

    We believe you should know how we're incentivized:

    Report Sales

    Our primary revenue comes from selling environmental screening reports. This is the core of our business.

    Affiliate Program

    Real estate professionals can earn commissions by referring clients. Affiliate relationships are disclosed per FTC guidelines.

    What We Don't Do

    • ×We don't sell your data or location history to third parties
    • ×We don't accept payment to include or exclude facilities from reports
    • ×We don't have advertising relationships with remediation companies
    • ×Our scores are not influenced by commercial relationships

    Questions about our approach?

    We're committed to transparency. If you have questions about our practices, we'd like to hear from you.