DeepSensi™

Press & media

The story, the numbers,
and everything you may print.

DeepSensi™ is the SIL-4 Cognitive Medical OS: a complete operating system for clinical intelligence, operational today, with a certified SIL-4-grade safety floor and results at the top of the field on medicine’s hardest benchmark. Everything on this page is fact-checked and free to publish.

Boilerplate

Copy-ready. Two lengths.

Short (about 60 words)
DeepSensi™ is the SIL-4 Cognitive Medical OS: a complete operating system for clinical intelligence. Its consilium engine reaches 86.0% top-1 accuracy on the NEJM CPC benchmark, where unaided physicians reach 19.9%, with a 0.0% missed-critical rate and a certified SIL-4-grade bound on hallucination. Operational in version 90, patent-pending, deployable fully air-gapped.
Full (about 120 words)
DeepSensi™ is the SIL-4 Cognitive Medical OS: a complete, decentralized operating system for clinical intelligence. A consilium of specialist AI counterparts for every discipline of medicine deliberates on each case under independent judges and a physician’s final word, reaching 86.0% top-1 and 93.7% top-3 on the NEJM CPC benchmark, where unaided physicians reach 19.9%, with a 0.0% missed-critical rate. Every conclusion clears a certified SIL-4-grade bound on hallucination and is audited to the last assertion; when evidence is missing, the LIMBO protocol declares “I don’t know.” The platform spans a physician workspace, lifelong patient memory, a royalty-paying knowledge economy, the GERN public-health radar, and sovereign, air-gapped deployment. Governed by the open, royalty-free DeepSensi™ Standard. Architected by Tomasz Jan Gomoła, Founder & Chief Architect, DeepSensi PBC, Dover, DE.

Fact sheet

The numbers, sourced.

Category
SIL-4 Cognitive Medical OS, a category DeepSensi defines: full-stack clinical intelligence with an engineered safety case.
Status
Operational, version 90, infrastructure on six continents. Not a prototype.
Accuracy
86.0% top-1, 93.7% top-3 on 301 NEJM Clinicopathological Conference cases (2014 to 2023), development-cohort; unaided physicians reach 19.9% on the same class of cases (SDBench, Nori et al. 2025).
Safety
0.0% missed-critical rate; certified worst-case hallucination bound of 3.23 × 10⁻⁶ per assertion, meeting the IEC 61508 SIL-4 demand-mode target with an approximately 31 times margin (WP-001).
Refusal
LIMBO, the structured “I don’t know”: when evidence is insufficient the system withholds and names the single test that settles the case.
Standard
The DeepSensi Standard (Gomola Framework, DSS-001): four certification levels, five auditable pillars, open and royalty-free for the entire industry.
Economy
The first economy for clinical expertise: outcome-gated physician royalties, a cryptographic reputation score, and a hypothesis marketplace with proof of authorship.
Sovereignty
Four deployment tiers up to a fully air-gapped hospital node running the complete consilium on premises; satellite-degraded field mode.
Public health
GERN: de-identified clinical notes become DeepSensi’s exclusive epidemic radar, surfacing clusters weeks before official dashboards, royalty-free for public-health bodies (WP-007).
Intellectual property
Twelve patent applications pending, twenty-four further in preparation. Trademark filed for DeepSensi™.
Founder
Tomasz Jan Gomoła, Founder & Chief Architect (ORCID 0009-0001-5222-6154). Portrait included in the press kit.
Company
DeepSensi PBC, a Public Benefit Corporation, 8 The Green STE A, Dover, DE 19901, USA.

Development-cohort figures are reported exactly as in the technical papers; a confirmatory prospective study is designated. We publish the proof, the standard, and the audit. We do not publish the blueprint.

Quotable

“Medicine has been asked to trust AI on faith. We refuse the premise. DeepSensi ships with the proof, the standard, and the audit.”

Also attributable to Tomasz Jan Gomoła, Founder & Chief Architect: “The rarest words in medical AI are ‘I don’t know.’ We made them a clinical output, and that is why nothing critical slips through.” · “Every other AI extracts physician expertise for free. We built the first one that pays for it.”

Story angles

Eight ways in. Pick yours.

The breakthrough

The machine that out-diagnoses the field

86.0% on the cases that defeat physicians, ahead of Microsoft’s MAI-DxO, with a safety case no competitor claims.

The refusal

The first AI honest enough to say “I don’t know”

LIMBO makes uncertainty a clinical output. Zero missed critical diagnoses is the consequence, not the slogan.

The economy

The AI that pays doctors

Outcome-gated royalties, a hypothesis marketplace, reputation as capital: the first economy for clinical expertise.

The radar

DeepSensi’s exclusive epidemic radar, made of paperwork

GERN turns ordinary de-identified notes into early warning, weeks ahead of dashboards, royalty-free for public health.

The sovereignty

The hospital that never sends a byte

The complete consilium runs fully air-gapped, from national systems to field deployments on satellite-degraded links.

The memory

Medicine that remembers you

A digital twin keeps watching after the visit; when new evidence diverges from the expected trajectory, the case reopens itself.

The standard

A safety standard given away for free

The Gomola Framework brings aviation-grade certification to clinical AI, royalty-free, including for competitors.

The founder

One architect, thirty-six patent filings

Conceived, engineered, and formally validated by Tomasz Jan Gomoła; papers, standard, and platform published together.

Read online

The release and five features, ready to quote.

Each piece is publication-ready and may be reprinted with attribution to DeepSensi™ PBC.

In the kit

Ready-to-send, fact-checked, free to use.

Press release

The announcement, print-ready, in DOCX and PDF.

Six feature stories

The breakthrough, the physician economy, the epidemic radar, medicine that remembers, and more.

Fact sheet & quotes

The numbers above, the boilerplate, and attributable founder quotes.

Brand assets & portrait

Logo, wordmark, icon, social card, and the founder’s portrait.

Media enquiries: [email protected]. Interviews with the founder available on request. Technical fact-checking against the papers: see the publications library.