Valkyrie Holmes on Faura and Property Resilience

What Valkyrie Holmes Taught Venture Step About Property Resilience

Valkyrie Holmes came onto Venture Step with an idea that sounded harsh and became more generous as she explained it: "you're not special." Her point was not that individual work is meaningless. It was that meaningful work is not reserved for a separate class of people who were born ready.

That belief connected both halves of E059. One half concerned the physical condition of a property facing wildfire, flood, hail, or hurricane. The other concerned a founder entering insurance without treating the industry's language and institutions as inaccessible.

Both problems become more workable when the important details are visible. A homeowner needs to understand what can be changed. An underwriter needs current information and a defined decision. A founder needs a map of the people, language, rules, and evidence that shape the market.

From suppression to mitigation

Valkyrie described an earlier wildfire project built around drones and vortex cannons. The concept aimed to contain a fire while reducing risk to frontline workers. It was memorable technology, but the lesson she carried forward was less cinematic.

She concluded that suppression did not reach the root of the problem she wanted to solve. Too much attention arrived after a fire had already started. Mitigation and the condition of the built environment offered another place to intervene.

That change in focus eventually led to Faura. In the episode, Valkyrie described meeting co-founder Amanda Southworth through a wildfire community online, sketching an early product, and incorporating before the two had met in person. Faura's current About page says the company was founded in 2023 and identifies Holmes as CEO and Southworth as CTO.

The company now describes itself differently from the episode's early emphasis on a single survivability score. Its current site presents a broader loop: generate property intelligence, engage policyholders, verify completed mitigation, and return updated first-party information to insurance teams.

That is Faura's description of its product. It is not independent confirmation that the resulting information predicts a specific loss, changes an underwriting decision, or improves an insurance outcome.

Hazard was only part of the picture

The insurance discussion started with a useful distinction. A hazard model can describe the likelihood and intensity of an event in a place. It does not, by itself, tell an insurer how a particular structure will perform.

The NAIC's catastrophe-model overview describes hazard, vulnerability, exposure, and financial modules. Vulnerability connects event intensity with expected damage based on building characteristics. Exposure includes location, risk characteristics, insured values, and policy terms. Financial outputs then reflect deductibles, limits, attachment points, and reinsurance.

Valkyrie used the word survivability for the structure-level question. In the episode, she imagined two properties exposed to a similar hazard but built and maintained differently. One might have better defensible space, roof characteristics, vents, materials, or other conditions that affect damage.

The example helped Dalton see why broad exposure and property condition should not be collapsed. It did not prove that any particular score represented those conditions accurately.

flowchart LR
    A["Hazard in a place"] --> B["Property conditions"]
    B --> C["Possible physical damage"]
    C --> D["Coverage and financial terms"]
    D --> E["Insurance outcome"]

The accuracy claim remained open

Late in the interview, Valkyrie said Faura could predict survivability with 95 percent accuracy. That statement made the buyer's central question more urgent, not less.

Accuracy has no stable meaning without the predicted outcome, peril, event set, population, prevalence, threshold, sampling method, holdout design, comparator, error distribution, and evaluation date. A model could appear highly accurate by predicting the common outcome while missing the cases that matter most.

Faura later published company-reported results involving selected wildfire burn scars. Those materials remain vendor evidence until the population, labels, exclusions, sampling, leakage controls, model version, uncertainty, and independent validation can be reviewed.

E059 should therefore be read as a record of a product thesis and an important procurement question. It is not Venture Step's validation of the model.

Learning insurance in public

The second durable idea came when Dalton asked how someone from outside insurance learned to speak credibly with its practitioners.

Valkyrie described starting with conversations across many roles, then reading, listening, taking notes, and reusing the language she heard. Sometimes another person corrected her. She treated that correction as part of the method.

She also explained why public synthesis mattered. Turning a conversation into plain-language content forced her to decide whether she actually understood it. The posts then brought more conversations, including with people who disagreed or supplied missing context.

This is a stronger model than merely collecting contacts. Each conversation produces a provisional claim. The founder researches it, explains it, invites correction, compares it with real workflows and rules, and updates the record.

The risk is obvious. Posting frequently can spread a mistake faster than it finds a correction. Confidential information, personal data, regulated records, and customer claims cannot be turned into content simply because they were heard in an interview. The founder still needs qualified people to own decisions that require a license, filing, technical standard, or professional judgment.

The shared idea was actionability

The product and founder stories met at the same point.

A broad warning that a place is risky may be too general to guide a property owner. A score without drivers, uncertainty, and a permitted use may be too opaque to guide an insurer. A founder who learns insider vocabulary without learning the decisions behind it may sound fluent while remaining wrong.

The better question is what a person can responsibly do next. That requires more than information. It requires a defined decision, understandable reasons, realistic actions, evidence that the action occurred, and a later check of what changed.

The Federal Insurance Office's nationwide analysis shows why this work matters. Its 2018 through 2022 dataset linked higher expected climate-related losses with higher premiums and higher nonrenewal rates across ZIP Codes. That is market context, not proof that one property intervention changes coverage.

The NAIC's national resilience strategy likewise puts pre-disaster mitigation, public risk information, catastrophe modeling, and solvency tools in the same policy conversation. Implementation remains state based and peril specific.

E059 did not solve those questions. It made them legible enough to ask more precisely. That is a meaningful starting point.

Listen to the original conversation

The March 2025 episode is available on Spotify and YouTube. The recording preserves what Valkyrie and Dalton said at the time. It does not establish current product behavior, company results, model accuracy, regulation, or insurance outcomes.

Editorial and AI disclosure

This article was developed from the preserved E059 transcript with AI assistance for research organization, drafting, and editing. Dalton Anderson remains the named author and must approve the final text. Valkyrie Holmes should review her identity, history, attributed ideas, and links before publication. Insurance, actuarial, catastrophe-model, engineering, legal, regulatory, and product claims require the review described in the linked package records.

This draft is not authorized for publication. It is not insurance, underwriting, pricing, coverage, engineering, safety, legal, regulatory, investment, or product advice.