Company

Dedication to mission

We founded NeoAmorfic to give decisions made under uncertainty a reference that does not move: a reading of the physical state of the system in question, derived by the laws of physics from the system’s own record and reproducible by anyone who holds the same data. The world is about to make more of its decisions with the help of systems that propose fluently and cannot say what is true. We believe the answer is not to trust them less but to check them better, and that the only check neither a machine nor a person can argue with is physics. NeoAmorfic is a deep-technology company in the literal sense: what it offers rests on physics that has to be proved, not on software alone.

We work in domains where close enough is catastrophic, and we accept the standard that comes with them. Nothing we publish rests on assurance; every reading can be re-run, every study is pre-registered, and every result, including the ones that went against us, is kept on the record. The company is founder-led and independent by design, and it will stay that way while the physics-intelligence layer is built.

We have started with financial markets and biology, the two domains in which a genuine record was public and the ground truth already established. Our goal is to extend the same layer to space and aerospace, to climate and energy systems, to industrial machinery and to autonomous systems — to any system that leaves a genuine record of its motion and in which close enough is catastrophic. Each new domain is admitted the way the first two were: a certified measurement, a named state and a public record, or nothing.

Our responsibility

We measure; we do not advise

NeoAmorfic manages no capital, executes no transactions and gives no personalised advice. A reading is information about the state of a system. The decision, and responsibility for its consequences, remain with the client, and we design every product so that this division cannot be blurred.

We say what a reading is not

A reading is not a forecast, and we do not describe it as one. Historical figures are backtests on real data and are labelled as such; only the sealed forward record is live. A reading is an occasion to act within a client’s own policy limits, and its contribution is judged across many decisions rather than any single one. When an incident affects the record, it is published with its cause.

We claim physics only where it is real

The engine is applied only to data that is a genuine record of a system over time. Where our own audits have found that a product did not rest on the physics, the claim has been withdrawn and the product removed from this account of the company. The name physics intelligence is earned per domain, not assumed.

We keep the negative results

Acceptance criteria are fixed before data is examined. A study that fails them is recorded as a result, not re-run until it passes, and it stays on the record beside the studies that succeeded. A reader who wants to know what the engine cannot yet do will find it written down.

We make no clinical claim

Our biology platform runs no wet laboratory and claims nothing about patients. Its purpose is to recover what biology has established and then to look for what established methods miss, in collaboration with the scientists who own the data and the questions.

We protect what we have built without hiding how it is checked

We describe the families of physical measurement the engine computes and the certification it has passed; we do not publish the constructions. Anyone can verify the record without the code, because every reading is committed before it is revealed and can be reconciled against independent recomputation.

How we work

Nothing is accepted on assurance

A claim exists when it can be re-run independently from sealed inputs and reproduces to the digit. Labels, reports and confident summaries are not evidence.

Studies are pre-registered

Scope, settings and acceptance criteria are frozen and hashed before a calculation is run. A result that fails them is recorded as a result.

Readings are committed before they can be checked

Every reading is written to a hash chain before it is revealed, so that anyone can verify afterwards that it was not changed. Where a product publishes a public feed, the feed shows the record with a delay.

The engine is sealed

The certified code sits under a hash manifest, and products reach it only through a gateway that refuses to run if anything has drifted. A change to the engine is a new certification, never a quiet update.

NeoAmorfic™ Ltd

NeoAmorfic Ltd is registered in England and Wales and was incorporated in November 2024. It is founder-led and independent: the engine’s architecture, its verification rules and every product decision are those of its founder, Michael Anton J. Tupay. The company works from London and Athens.

Registered office: 71 Shelton Street, Covent Garden, London WC2H 9JQ, United Kingdom
Institutional enquiries: institutional@neoamorfic.ai
Founder: vision@neoamorfic.ai