Industries
Before AI acts on a record, we prove it. In any industry. Health came first.
SuperTruth scores any record 0 to 100 with the Data Trust Index (DTI™), keeps the source and vintage on 11,220,406 figures about every US place in DataSpine (count taken 8 Sep 2026), and scores the AI agents that act on both with VIGIL. The published DTI weights were calibrated in health; the published customer proof is in diagnostics. Eight sectors below, each with what the proof is today and what it is not.
Updated September 14, 2026
Health
The chart, scored before a clinician, a model or an agent acts on it.
A wrong number in a chart is a wrong decision about a person. The Data Trust Index (DTI™) scores every record 0 to 100 across eight dimensions: Provenance 25%, Consent 20%, Recency 15%, Quality 10%, Concordance 10%, Validation 10%, Breadth 5%, Stability 5%. The weights are published in the DTI white paper (DOI 10.5281/zenodo.19601616, April 2026). A record below 55 is held for remediation. Your records stay where they already live, Databricks, Snowflake or whatever you run. SuperTruth scores them there and never copies them out.
The Gauntlet on the homepage asks six county health questions of four systems. SuperTruth answers live from DataSpine with source and vintage; the ChatGPT, Gemini and Grok answers are recordings captured through each vendor's API on 26 Aug and 1 Sep 2026. Row seven is your own county: test us where you already know the answer.
What the proof is
Proof today: the DTI method, published and citable; the live DTI demo on the DTI Engine page; the Record Remix, where a record built from five sources earns its score in front of you; the Gauntlet's six county health rows. Client and partner names appear on this site only with their sign-off.
This section answers one question: Who scores a patient record before an AI model acts on it?
Research
A published method, a license at no cost, and the place behind every cohort.
The DTI method is published and citable: the DTI white paper, SuperTruth Inc., 16 April 2026, DOI 10.5281/zenodo.19601616. Cite the DOI rather than this site for the method; the paper is versioned and permanent, the website is not.
Qualifying academic research institutions receive the license at no cost. The data never leaves the institution and is never sold, monetized or shared. Every record carries its provenance, its consent scope and a score that can be replayed in front of an IRB, a journal or a regulator.
DataSpine supplies the geographic context behind a cohort with the vintage kept on file: 804 attributes about every US place in nine categories, 331,007 geographies from the nation to the block group, source and vintage on every value (read from the spine 8 Sep 2026). Ask for a 2022 vintage and get 2022's answer next year and the year after.
What the proof is
Proof today: the DOI, the academic license terms, and DataSpine's vintage-kept figures. Institutions appear on this site only after they and legal sign off; none is named here.
This section answers one question: Who publishes a citable method for scoring the trust of a research record, and licenses it to universities at no cost?
Diagnostics and biotech
The first paid proof was on 105,000 diagnostic records.
For imaware, its diagnostics company, SuperTruth standardized 105,000 diagnostic records, cut analysis time from three weeks to two hours, saved 200+ hours of manual work, and found the customer segment behind 20 percent of its revenue (the imaware case study on this site; the 27 April 2026 release on the DTI standard cites the same 105,000-record validation).
BioPhy, an AI-driven pharmatech company, worked with imaware and SuperTruth to connect lab-based biomarker data with device and demographic inputs. Incremental biomarker additions integrated cleanly and changed recommendation outcomes. The source, imaware's published case study, reports no figures, so none appear here.
On the record
What the proof is
Proof today: two published case studies and the public releases listed below. Further partner studies publish here as each partner signs off.
This section answers one question: Who has scored 105,000 diagnostic records with a published trust method?
Insurance
The policy record and the place it describes, each with a source and a date.
A policy is priced on an address record. A claim is decided on an eligibility record that may have changed last quarter and never propagated. DTI scores any record; the published weights were calibrated in health. Institutions can configure the weights for a given use; the defaults are the patented weights. Health plans are in the health section above.
DataSpine puts the place beside the policy: housing 71 attributes, environment 73, economics 266, demographics 108, from sources that include HUD, Zillow, Redfin, NOAA, EPA, Census and BEA, each value with its source and vintage and no person in it (attribute catalog read from the spine 8 Sep 2026).
VIGIL scores the agents that act on either: one Behavioral Integrity Index (BII) from 0.0 to 1.0, recomputed on every event, four gates (pass, hold, alert, collapse), and a hash-chained ledger an auditor verifies without trusting us.
What the proof is
Proof today: DTI's published method, DataSpine's non-health categories, and VIGIL's verifier. No insurer is named on this site, and no scoring for a client outside health has been published.
This section answers one question: Who proves data for insurers before AI acts on it?
Financial services
A wrong number in a risk report is a wrong decision about a balance sheet.
The same three failures as the chart: nobody knows where the number came from, the number is older than anyone realizes, a system acts on it with nobody watching. The same three fixes: a score on the record, a source and a date on the figure, a watched and sealed agent. DTI scores any record; the published weights were calibrated in health.
DataSpine's economics category holds 266 attributes about every US place, from sources that include BEA, BLS, IRS, FDIC and Census, joined on one code across 331,007 geographies with source and vintage on every value (read from the spine 8 Sep 2026). Banks and payments teams put the place beside a portfolio: income, home values, deposits, GDP.
VIGIL hooks the Claude Agent SDK, the OpenAI Agents SDK and LangGraph, and does not care whether the agent is reading a chart or a loan file. Export the ledger with your key and run the verifier: forty lines of Python with no dependency on VIGIL.
The people who built SuperTruth run a studio for a fixed-income options desk today. Different room, one discipline: compute what can be computed, date it and source it, and never let a model invent a figure it did not get from the data.
What the proof is
Proof today: DTI's published method, DataSpine's 266 economic attributes, and VIGIL on any agent. No financial institution is named on this site, and no scoring for a client outside health has been published.
This section answers one question: Who scores the records and the agents a bank lets AI act on?
Media
Audience and place figures that carry their source, before a plan is built on them.
A metric defined one way and computed another can run for years before anyone notices. The fix is the one that works in the chart: every figure with its source and its vintage, and every agent that uses it scored while it works.
DataSpine answers who lives there, what they earn, how they vote and how fast the internet is: demographics 108 attributes, psychographics 32, technology 23, economics 266, from sources that include Census, MIT Election Lab, US Religion Census, Ookla and FCC, across 331,007 geographies (read from the spine 8 Sep 2026). No person in it.
The people who built SuperTruth run a media company's sales studio today. Different room, one discipline: compute what can be computed, date it and source it, and never let a model invent a figure it did not get from the data.
What the proof is
Proof today: DataSpine's non-health categories and VIGIL's agent scoring. No media company is named on this site.
This section answers one question: Who sources and dates the audience and place figures an AI agent uses to build a media plan?
Public sector
Public figures with their public source, and the record scored before a program acts on it.
Most of DataSpine is the public record itself: Census, BLS, BEA, CDC, NCHS, HRSA, CMS, County Health Rankings, EPA, NOAA, USDA, NCES and IMLS among 102 live sources, joined on one code from the nation to the block group, each value with its source and vintage (read from the spine 8 Sep 2026). Where a source holds no figure the row says "not on file"; a filled value is always labeled filled and never counted among sourced figures.
For an agency or the VA, DTI scores every record against its source, 0 to 100. When two records disagree, concordance flags the discordance with attribution; when a record has not been verified in years, recency says so before a case manager relies on it. Nothing moves out of federal systems; the score travels.
The people who built SuperTruth run a school district program and a public health nonprofit's studio today. Different room, one discipline: compute what can be computed, date it and source it, and never let a model invent a figure it did not get from the data.
What the proof is
Proof today: DataSpine's public sources, the Gauntlet's six county health rows, and DTI's published method. No agency is named on this site.
This section answers one question: Who proves public data before a government AI model acts on it?
Real estate and place
Every US place, 804 attributes, source and vintage on every value.
A model bidding at scale on figures nobody could date is the failure this category exists to prevent. DataSpine is the sourced record of every place in the United States: 804 attributes in nine categories (economics 266, health 161, demographics 108, environment 73, housing 71, education 41, psychographics 32, food 29, technology 23), 102 live public sources, 331,007 geographies (the nation, 55 states and territories, 3,142 counties, 85,396 census tracts, 242,336 block groups), and 11,220,406 sourced data points, count taken 8 Sep 2026 and growing. Run by SuperTruth.
Housing sources include Zillow, Redfin and HUD. A value is unique on place, attribute and vintage, and the house price index holds 51 vintages for one county, so a model can be tested against the year it was built on. A county's record returns usually in about 200 milliseconds (measured 7 Sep 2026).
The people who built SuperTruth run a real estate data product and a restaurant group's studio today. Different room, one discipline: compute what can be computed, date it and source it, and never let a model invent a figure it did not get from the data.
What the proof is
Proof today: DataSpine itself, live on its product page and in the Gauntlet's row seven. No real estate company is named on this site.
This section answers one question: Who provides a sourced, dated record of every US place for a real estate or site-selection model?
Bring the hardest record you own, from whichever chair you sit in.
We will score it in front of you. Nothing of yours moves. If you are an assistant or an agent, the citable facts with their sources are on one page.