Portfolio
MOBLUEHQ builds AI verification infrastructure — systems that prove how an AI result was produced, and where. We are building toward running more of that work on ordinary machines instead of hyperscaler data centers. Our first products release soon; our research spans the regulated industries where a wrong answer is expensive.
Product pages are password protected until that product launches or opens for demos. If you have a password, use Enter. If not, request access — a person reads every request. Behind the gate is the real product: what it does, how to set it up, and what it is not.
First releases
blueHealthcare
The record that defends the decision. Verification tools for utilization management, claims, clinical-trial source data and lab integrity, on a shared clinical platform. Each one leaves a signed receipt anyone can check offline, and a named human always decides.
Not a diagnostic tool, and never the decision-maker. A clinician decides; this records how.
blueMonster
The idea-discovery engine for Mac. Ask it a hard question; overnight it generates, attacks and develops candidates, then hands you an evidence dossier with signed, receipt-backed runs and MBR export.
Not a source of truth. It generates hypotheses — verify before you rely on any of them.
Then, in this order
The next three. Position reflects the actual release sequence -- each is built to make the fleet support a different shape of product, which is why they ship in this order rather than by readiness.
A calm, private AI assistant for your Mac. Runs on-device by default; escalate on your own terms, never automatically.
Not an auto-escalating router — you choose when a question goes past the on-device model.
Second-pass prior-art search. Screens filings for relevant prior art and anchors each quote to its exact place in the source, or refuses to show it. Delivered as a service; every run produces a tamper-evident audit pack you can verify offline in your browser.
Not a certification that no prior art exists. It is a second look at what your existing tool may have missed.
Find claims the recorded evidence and tools could not support. Reads a document against the evidence behind it and marks every claim as supported, unsupported, unsupportable, or unresolved.
Not a global fact checker or legal opinion.
After the first wave
Named, with no date and no capability claim, because nothing here is ready to promise.
blueFloor
blueIntent
blueParity
bluePipelineThe layer underneath
Not sold separately. Everything above runs on these.
The verification substrate every MOBLUEHQ product runs on. A private inference mesh that returns sealed, signed results — the scheduler never sees your data, and every answer carries a receipt anyone can verify.
A receipt records process, not truth.
The compute mesh we're building toward — pool a machine's spare capacity to help run verified AI work. Not a public download yet: today's build is internal and unsigned, run only on our own machines. This is the direction we're building toward, not what happens today. Much of the compute the world needs, does not need a massive data center.
Areas of interest & research
Financial AI Verification
Evidence and provenance for AI used in credit, risk, and settlement — where regulators and counterparties demand a checkable trail, not a black box.
Legal AI Verification
Auditability for AI in contracts, disputes, and compliance — process integrity a court or opposing party could independently validate.
Manufacturing AI Verification
Trustworthy autonomy for industrial control and inspection — where an unverified action changes the physical world and cannot simply be undone.